MOLECULAR QUANTUM MECHANICS, FOURTH EDITION
Peter Atkins Ronald Friedman
OXFORD UNIVERSITY PRESS
MOLECULAR QUANTUM M...
146 downloads
1349 Views
6MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
MOLECULAR QUANTUM MECHANICS, FOURTH EDITION
Peter Atkins Ronald Friedman
OXFORD UNIVERSITY PRESS
MOLECULAR QUANTUM MECHANICS
This page intentionally left blank
MOLECULAR QUANTUM MECHANICS FOURTH EDITION
Peter Atkins University of Oxford Ronald Friedman Indiana Purdue Fort Wayne
AC
AC
Great Clarendon Street, Oxford OX2 6DP Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Bangkok Buenos Aires Cape Town Chennai Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi Sa˜o Paulo Shanghai Taipei Tokyo Toronto Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York #
Peter Atkins and Ronald Friedman 2005
The moral rights of the authors have been asserted. Database right Oxford University Press (maker) First published 2005 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose this same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available ISBN 0--19--927498--3 10 9 8 7 6 5 4 3 2 1 Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India Printed in Great Britain on acid-free paper by Ashford Colour Press
Table of contents Preface Introduction and orientation 1 The foundations of quantum mechanics 2 Linear motion and the harmonic oscillator 3 Rotational motion and the hydrogen atom
xiii 1 9 43 71
4 5 6 7 8
Angular momentum Group theory Techniques of approximation Atomic spectra and atomic structure An introduction to molecular structure
9 10 11 12
The calculation of electronic structure Molecular rotations and vibrations Molecular electronic transitions The electric properties of molecules
287
13 The magnetic properties of molecules 14 Scattering theory Further information Further reading
436
Appendix 1 Character tables and direct products Appendix 2 Vector coupling coefficients Answers to selected problems Index
557
98 122 168 207 249 342 382 407 473 513 553 562 563 565
This page intentionally left blank
Detailed Contents Introduction and orientation
1
The plausibility of the Schro¨dinger equation
36
1.22 The propagation of light
36
0.1 Black-body radiation
1
1.23 The propagation of particles
38
0.2 Heat capacities
3
1.24 The transition to quantum mechanics
39
0.3 The photoelectric and Compton effects
4
0.4 Atomic spectra
5
0.5 The duality of matter
6
PROBLEMS
40
PROBLEMS
8
2 Linear motion and the harmonic oscillator
43
1 The foundations of quantum mechanics
9
The characteristics of acceptable wavefunctions
43
Some general remarks on the Schro¨dinger equation
44
Operators in quantum mechanics
9
2.1 The curvature of the wavefunction
45
1.1 Linear operators
10
2.2 Qualitative solutions
45
1.2 Eigenfunctions and eigenvalues
10
2.3 The emergence of quantization
46
1.3 Representations
12
2.4 Penetration into non-classical regions
46
1.4 Commutation and non-commutation
13
1.5 The construction of operators
14
1.6 Integrals over operators
15
1.7 Dirac bracket notation
16
1.8 Hermitian operators
17
The postulates of quantum mechanics 1.9 States and wavefunctions
19 19
1.10 The fundamental prescription
20
1.11 The outcome of measurements
20
1.12 The interpretation of the wavefunction
22
1.13 The equation for the wavefunction
23
1.14 The separation of the Schro¨dinger equation
23
The specification and evolution of states
25
Translational motion
47
2.5 Energy and momentum
48
2.6 The significance of the coefficients
48
2.7 The flux density
49
2.8 Wavepackets
50
Penetration into and through barriers 2.9 An infinitely thick potential wall
51 51
2.10 A barrier of finite width
52
2.11 The Eckart potential barrier
54
Particle in a box
55
2.12 The solutions
56
2.13 Features of the solutions
57
2.14 The two-dimensional square well
58
2.15 Degeneracy
59
1.15 Simultaneous observables
25
1.16 The uncertainty principle
27
1.17 Consequences of the uncertainty principle
29
1.18 The uncertainty in energy and time
30
2.16 The solutions
61
1.19 Time-evolution and conservation laws
30
2.17 Properties of the solutions
63
2.18 The classical limit
65
Matrices in quantum mechanics
32
The harmonic oscillator
60
1.20 Matrix elements
32
Translation revisited: The scattering matrix
66
1.21 The diagonalization of the hamiltonian
34
PROBLEMS
68
viii
j
CONTENTS
3 Rotational motion and the hydrogen atom
71
The angular momenta of composite systems 4.9 The specification of coupled states
Particle on a ring
71
3.1 The hamiltonian and the Schro¨dinger equation
71
3.2 The angular momentum
73
3.3 The shapes of the wavefunctions
74
3.4 The classical limit Particle on a sphere
76 76
3.5 The Schro¨dinger equation and its solution
76
3.6 The angular momentum of the particle
79
3.7 Properties of the solutions
81
3.8 The rigid rotor
82
Motion in a Coulombic field 3.9 The Schro¨dinger equation for hydrogenic atoms
112 112
4.10 The permitted values of the total angular momentum
113
4.11 The vector model of coupled angular momenta
115
4.12 The relation between schemes
117
4.13 The coupling of several angular momenta
119
PROBLEMS
120
5 Group theory
122
The symmetries of objects
122
5.1 Symmetry operations and elements
123
5.2 The classification of molecules
124
84 The calculus of symmetry
129
84
5.3 The definition of a group
129
3.10 The separation of the relative coordinates
85
5.4 Group multiplication tables
130
3.11 The radial Schro¨dinger equation
85
5.5 Matrix representations
131
5.6 The properties of matrix representations
135
3.12 Probabilities and the radial distribution function
90
5.7 The characters of representations
137
3.13 Atomic orbitals
91
5.8 Characters and classes
138
3.14 The degeneracy of hydrogenic atoms
94
5.9 Irreducible representations
139
PROBLEMS
96
5.10 The great and little orthogonality theorems Reduced representations
4 Angular momentum
98
The angular momentum operators
98
4.1 The operators and their commutation relations
99
142 145
5.11 The reduction of representations
146
5.12 Symmetry-adapted bases
147
The symmetry properties of functions
151
5.13 The transformation of p-orbitals
151
5.14 The decomposition of direct-product bases
152
4.2 Angular momentum observables
101
5.15 Direct-product groups
155
4.3 The shift operators
101
5.16 Vanishing integrals
157
5.17 Symmetry and degeneracy
159
The definition of the states
102
4.4 The effect of the shift operators
102
4.5 The eigenvalues of the angular momentum
104
4.6 The matrix elements of the angular momentum
106
4.7 The angular momentum eigenfunctions
108
4.8 Spin
110
The full rotation group
161
5.18 The generators of rotations
161
5.19 The representation of the full rotation group
162
5.20 Coupled angular momenta
164
Applications PROBLEMS
165 166
CONTENTS
6 Techniques of approximation
168
j
ix
7.10 The spectrum of helium
224
7.11 The Pauli principle
225
Time-independent perturbation theory
168
6.1 Perturbation of a two-level system
169
7.12 Penetration and shielding
229
6.2 Many-level systems
171
7.13 Periodicity
231
6.3 The first-order correction to the energy
172
7.14 Slater atomic orbitals
233
6.4 The first-order correction to the wavefunction
174
7.15 Self-consistent fields
234
6.5 The second-order correction to the energy
175
6.6 Comments on the perturbation expressions
176
7.16 Term symbols and transitions of many-electron atoms
236
6.7 The closure approximation
178
7.17 Hund’s rules and the relative energies of terms
239
6.8 Perturbation theory for degenerate states
180
7.18 Alternative coupling schemes
240
Variation theory 6.9 The Rayleigh ratio 6.10 The Rayleigh–Ritz method
183
7.19 The normal Zeeman effect
242
7.20 The anomalous Zeeman effect
243
7.21 The Stark effect
245
Time-dependent perturbation theory
189
6.11 The time-dependent behaviour of a two-level system
189
6.12 The Rabi formula
192
6.13 Many-level systems: the variation of constants
193
6.14 The effect of a slowly switched constant perturbation
195
6.15 The effect of an oscillating perturbation
197
6.16 Transition rates to continuum states
199
6.17 The Einstein transition probabilities
200
6.18 Lifetime and energy uncertainty
203
The spectrum of atomic hydrogen
242
185 187
7 Atomic spectra and atomic structure
Atoms in external fields
229
183
The Hellmann–Feynman theorem
PROBLEMS
Many-electron atoms
204
207
207
PROBLEMS
246
8 An introduction to molecular structure
249
The Born–Oppenheimer approximation
249
8.1 The formulation of the approximation
250
8.2 An application: the hydrogen molecule–ion
251
Molecular orbital theory
253
8.3 Linear combinations of atomic orbitals
253
8.4 The hydrogen molecule
258
8.5 Configuration interaction
259
8.6 Diatomic molecules
261
8.7 Heteronuclear diatomic molecules
265
Molecular orbital theory of polyatomic molecules
266
7.1 The energies of the transitions
208
7.2 Selection rules
209
8.8 Symmetry-adapted linear combinations
266
7.3 Orbital and spin magnetic moments
212
8.9 Conjugated p-systems
269
7.4 Spin–orbit coupling
214
8.10 Ligand field theory
7.5 The fine-structure of spectra
216
8.11 Further aspects of ligand field theory
7.6 Term symbols and spectral details
217
7.7 The detailed spectrum of hydrogen
218
The band theory of solids
274 276 278
8.12 The tight-binding approximation
279
The structure of helium
219
8.13 The Kronig–Penney model
281
7.8 The helium atom
219
8.14 Brillouin zones
284
7.9 Excited states of helium
222
PROBLEMS
285
x
j
CONTENTS
9 The calculation of electronic structure
287
10.3 Rotational energy levels
345
10.4 Centrifugal distortion
349
288
10.5 Pure rotational selection rules
349
288
10.6 Rotational Raman selection rules
351
9.2 The Hartree–Fock approach
289
10.7 Nuclear statistics
353
9.3 Restricted and unrestricted Hartree–Fock calculations
291
9.4 The Roothaan equations
The Hartree–Fock self-consistent field method 9.1 The formulation of the approach
The vibrations of diatomic molecules
357
293
10.8 The vibrational energy levels of diatomic molecules
357
9.5 The selection of basis sets
296
10.9 Anharmonic oscillation
359
9.6 Calculational accuracy and the basis set
301
10.10 Vibrational selection rules
360
302
10.11 Vibration–rotation spectra of diatomic molecules
362
9.7 Configuration state functions
303
9.8 Configuration interaction
303
10.12 Vibrational Raman transitions of diatomic molecules
364
9.9 CI calculations
Electron correlation
305
The vibrations of polyatomic molecules
365
9.10 Multiconfiguration and multireference methods
308
10.13 Normal modes
365
9.11 Møller–Plesset many-body perturbation theory
310
9.12 The coupled-cluster method
313
10.14 Vibrational selection rules for polyatomic molecules
368
10.15 Group theory and molecular vibrations
369
10.16 The effects of anharmonicity
373
Density functional theory
316
9.13 Kohn–Sham orbitals and equations
317
10.17 Coriolis forces
376
9.14 Exchange–correlation functionals
319
10.18 Inversion doubling
377
321
Appendix 10.1 Centrifugal distortion
379
PROBLEMS
380
Gradient methods and molecular properties 9.15 Energy derivatives and the Hessian matrix
321
9.16 Analytical derivatives and the coupled perturbed equations
322
Semiempirical methods
11 Molecular electronic transitions
382
The states of diatomic molecules
382
325
9.17 Conjugated p-electron systems
326
9.18 Neglect of differential overlap
329
11.1 The Hund coupling cases
382
332
11.2 Decoupling and L-doubling
384
9.19 Force fields
333
11.3 Selection rules
386
9.20 Quantum mechanics–molecular mechanics
334
Molecular mechanics
Software packages for electronic structure calculations
336
PROBLEMS
339
10 Molecular rotations and vibrations
342
Vibronic transitions
386
11.4 The Franck–Condon principle
386
11.5 The rotational structure of vibronic transitions
389
The electronic spectra of polyatomic molecules
390
11.6 Symmetry considerations
391
11.7 Chromophores
391
342
11.8 Vibronically allowed transitions
393
10.1 Absorption and emission
342
11.9 Singlet–triplet transitions
395
10.2 Raman processes
344
The fate of excited species
396
344
11.10 Non-radiative decay
396
Spectroscopic transitions
Molecular rotation
CONTENTS
j
xi
11.11 Radiative decay
397
Magnetic resonance parameters
452
11.12 The conservation of orbital symmetry
399
13.11 Shielding constants
452
11.13 Electrocyclic reactions
399
13.12 The diamagnetic contribution to shielding
456
11.14 Cycloaddition reactions
401
13.13 The paramagnetic contribution to shielding
458
11.15 Photochemically induced electrocyclic reactions
403
13.14 The g-value
459
11.16 Photochemically induced cycloaddition reactions
404
13.15 Spin–spin coupling
462
PROBLEMS
406
13.16 Hyperfine interactions
463
13.17 Nuclear spin–spin coupling
467
PROBLEMS
471
12 The electric properties of molecules
407
The response to electric fields
407
12.1 Molecular response parameters
407
12.2 The static electric polarizability
409
12.3 Polarizability and molecular properties
411
12.4 Polarizabilities and molecular spectroscopy
413
12.5 Polarizabilities and dispersion forces
414
12.6 Retardation effects
418
Bulk electrical properties
418
12.7 The relative permittivity and the electric susceptibility
418
12.8 Polar molecules
420
12.9 Refractive index
422
14 Scattering theory
473
The formulation of scattering events
473
14.1 The scattering cross-section
473
14.2 Stationary scattering states
475
Partial-wave stationary scattering states
479
14.3 Partial waves
479
14.4 The partial-wave equation
480
14.5 Free-particle radial wavefunctions and the scattering phase shift
481
14.6 The JWKB approximation and phase shifts
484
14.7 Phase shifts and the scattering matrix element
486
Optical activity
427
14.8 Phase shifts and scattering cross-sections
488
12.10 Circular birefringence and optical rotation
427
14.9 Scattering by a spherical square well
490
12.11 Magnetically induced polarization
429
14.10 Background and resonance phase shifts
12.12 Rotational strength
431
14.11 The Breit–Wigner formula
494
434
14.12 Resonance contributions to the scattering matrix element
495
Multichannel scattering
497
14.13 Channels for scattering
497
14.14 Multichannel stationary scattering states
498
14.15 Inelastic collisions
498
14.16 The S matrix and multichannel resonances
501
PROBLEMS
492
13 The magnetic properties of molecules
436
The descriptions of magnetic fields
436
13.1 The magnetic susceptibility
436
13.2 Paramagnetism
437
13.3 Vector functions
439
13.4 Derivatives of vector functions
440
The Green’s function
502
13.5 The vector potential
441
14.17 The integral scattering equation and Green’s functions
502
14.18 The Born approximation
504
Magnetic perturbations
442
13.6 The perturbation hamiltonian
442
13.7 The magnetic susceptibility
444
13.8 The current density
447
13.9 The diamagnetic current density
450
Appendix 14.1 The derivation of the Breit–Wigner formula Appendix 14.2 The rate constant for reactive scattering
13.10 The paramagnetic current density
451
PROBLEMS
508 509 510
xii
j
CONTENTS
Further information
513
Classical mechanics
513
15 Vector coupling coefficients Spectroscopic properties
535 537
16 Electric dipole transitions
537
1 Action
513
17 Oscillator strength
538
2 The canonical momentum
515
18 Sum rules
540
3 The virial theorem
516
19 Normal modes: an example
541
4 Reduced mass
518 The electromagnetic field
543
Solutions of the Schro¨dinger equation
519
5 The motion of wavepackets
519
6 The harmonic oscillator: solution by factorization
521
Mathematical relations
547
7 The harmonic oscillator: the standard solution
523
22 Vector properties
547
8 The radial wave equation
525
23 Matrices
549
9 The angular wavefunction
526
Further reading
553
Appendix 1 Appendix 2 Answers to selected problems Index
557 562 563 565
10 Molecular integrals
527
11 The Hartree–Fock equations
528
12 Green’s functions
532
13 The unitarity of the S matrix
533
Group theory and angular momentum 14 The orthogonality of basis functions
534 534
20 The Maxwell equations
543
21 The dipolar vector potential
546
PREFACE Many changes have occurred over the editions of this text but we have retained its essence throughout. Quantum mechanics is filled with abstract material that is both conceptually demanding and mathematically challenging: we try, wherever possible, to provide interpretations and visualizations alongside mathematical presentations. One major change since the third edition has been our response to concerns about the mathematical complexity of the material. We have not sacrificed the mathematical rigour of the previous edition but we have tried in numerous ways to make the mathematics more accessible. We have introduced short commentaries into the text to remind the reader of the mathematical fundamentals useful in derivations. We have included more worked examples to provide the reader with further opportunities to see formulae in action. We have added new problems for each chapter. We have expanded the discussion on numerous occasions within the body of the text to provide further clarification for or insight into mathematical results. We have set aside Proofs and Illustrations (brief examples) from the main body of the text so that readers may find key results more readily. Where the depth of presentation started to seem too great in our judgement, we have sent material to the back of the chapter in the form of an Appendix or to the back of the book as a Further information section. Numerous equations are tabbed with www to signify that on the Website to accompany the text [www.oup.com/uk/ booksites/chemistry/] there are opportunities to explore the equations by substituting numerical values for variables. We have added new material to a number of chapters, most notably the chapter on electronic structure techniques (Chapter 9) and the chapter on scattering theory (Chapter 14). These two chapters present material that is at the forefront of modern molecular quantum mechanics; significant advances have occurred in these two fields in the past decade and we have tried to capture their essence. Both chapters present topics where comprehension could be readily washed away by a deluge of algebra; therefore, we concentrate on the highlights and provide interpretations and visualizations wherever possible. There are many organizational changes in the text, including the layout of chapters and the choice of words. As was the case for the third edition, the present edition is a rewrite of its predecessor. In the rewriting, we have aimed for clarity and precision. We have a deep sense of appreciation for many people who assisted us in this endeavour. We also wish to thank the numerous reviewers of the textbook at various stages of its development. In particular, we would like to thank Charles Trapp, University of Louisville, USA Ronald Duchovic, Indiana Purdue Fort Wayne, USA
xiv
j
PREFACE
Karl Jalkanen, Technical University of Denmark, Denmark Mark Child, University of Oxford, UK Ian Mills, University of Reading, UK David Clary, University of Oxford, UK Stephan Sauer, University of Copenhagen, Denmark Temer Ahmadi, Villanova University, USA Lutz Hecht, University of Glasgow, UK Scott Kirby, University of Missouri-Rolla, USA All these colleagues have made valuable suggestions about the content and organization of the book as well as pointing out errors best spotted in private. Many individuals (too numerous to name here) have offered advice over the years and we value and appreciate all their insights and advice. As always, our publishers have been very helpful and understanding. PWA, Oxford RSF, Indiana University Purdue University Fort Wayne June 2004
Introduction and orientation
0.1 Black-body radiation 0.2 Heat capacities 0.3 The photoelectric and Compton effects 0.4 Atomic spectra 0.5 The duality of matter
There are two approaches to quantum mechanics. One is to follow the historical development of the theory from the first indications that the whole fabric of classical mechanics and electrodynamics should be held in doubt to the resolution of the problem in the work of Planck, Einstein, Heisenberg, Schro¨dinger, and Dirac. The other is to stand back at a point late in the development of the theory and to see its underlying theoretical structure. The first is interesting and compelling because the theory is seen gradually emerging from confusion and dilemma. We see experiment and intuition jointly determining the form of the theory and, above all, we come to appreciate the need for a new theory of matter. The second, more formal approach is exciting and compelling in a different sense: there is logic and elegance in a scheme that starts from only a few postulates, yet reveals as their implications are unfolded, a rich, experimentally verifiable structure. This book takes that latter route through the subject. However, to set the scene we shall take a few moments to review the steps that led to the revolutions of the early twentieth century, when some of the most fundamental concepts of the nature of matter and its behaviour were overthrown and replaced by a puzzling but powerful new description.
0.1 Black-body radiation In retrospect—and as will become clear—we can now see that theoretical physics hovered on the edge of formulating a quantum mechanical description of matter as it was developed during the nineteenth century. However, it was a series of experimental observations that motivated the revolution. Of these observations, the most important historically was the study of blackbody radiation, the radiation in thermal equilibrium with a body that absorbs and emits without favouring particular frequencies. A pinhole in an otherwise sealed container is a good approximation (Fig. 0.1). Two characteristics of the radiation had been identified by the end of the century and summarized in two laws. According to the Stefan–Boltzmann law, the excitance, M, the power emitted divided by the area of the emitting region, is proportional to the fourth power of the temperature: M ¼ sT 4
ð0:1Þ
2
j
INTRODUCTION AND ORIENTATION
Detected radiation
Pinhole
The Stefan–Boltzmann constant, s, is independent of the material from which the body is composed, and its modern value is 56.7 nW m2 K4. So, a region of area 1 cm2 of a black body at 1000 K radiates about 6 W if all frequencies are taken into account. Not all frequencies (or wavelengths, with l ¼ c/n), though, are equally represented in the radiation, and the observed peak moves to shorter wavelengths as the temperature is raised. According to Wien’s displacement law, lmax T ¼ constant
Container at a temperature T Fig. 0.1 A black-body emitter can be simulated by a heated container with a pinhole in the wall. The electromagnetic radiation is reflected many times inside the container and reaches thermal equilibrium with the walls.
25
/(8π(kT )5/(hc)4)
20 15 10 5
ð0:2Þ
with the constant equal to 2.9 mm K. One of the most challenging problems in physics at the end of the nineteenth century was to explain these two laws. Lord Rayleigh, with minor help from James Jeans,1 brought his formidable experience of classical physics to bear on the problem, and formulated the theoretical Rayleigh–Jeans law for the energy density e(l), the energy divided by the volume, in the wavelength range l to l þ dl: 8pkT ð0:3Þ deðlÞ ¼ rðlÞ dl rðlÞ ¼ 4 l where k is Boltzmann’s constant (k ¼ 1.381 10 23 J K1). This formula summarizes the failure of classical physics. It suggests that regardless of the temperature, there should be an infinite energy density at very short wavelengths. This absurd result was termed by Ehrenfest the ultraviolet catastrophe. At this point, Planck made his historic contribution. His suggestion was equivalent to proposing that an oscillation of the electromagnetic field of frequency n could be excited only in steps of energy of magnitude hn, where h is a new fundamental constant of nature now known as Planck’s constant. According to this quantization of energy, the supposition that energy can be transferred only in discrete amounts, the oscillator can have the energies 0, hn, 2hn, . . . , and no other energy. Classical physics allowed a continuous variation in energy, so even a very high frequency oscillator could be excited with a very small energy: that was the root of the ultraviolet catastrophe. Quantum theory is characterized by discreteness in energies (and, as we shall see, of certain other properties), and the need for a minimum excitation energy effectively switches off oscillators of very high frequency, and hence eliminates the ultraviolet catastrophe. When Planck implemented his suggestion, he derived what is now called the Planck distribution for the energy density of a black-body radiator: 8phc ehc=lkT ð0:4Þ l5 1 ehc=lkT This expression, which is plotted in Fig. 0.2, avoids the ultraviolet catastrophe, and fits the observed energy distribution extraordinarily well if we take h ¼ 6.626 1034 J s. Just as the Rayleigh–Jeans law epitomizes the failure of classical physics, the Planck distribution epitomizes the inception of rðlÞ ¼
0
0
0.5
1.0 1.5 kT /hc
Fig. 0.2 The Planck distribution.
2.0
.......................................................................................................
1. ‘It seems to me,’ said Jeans, ‘that Lord Rayleigh has introduced an unnecessary factor 8 by counting negative as well as positive values of his integers.’ (Phil. Mag., 91, 10 (1905).)
0.2 HEAT CAPACITIES
j
3
quantum theory. It began the new century as well as a new era, for it was published in 1900.
0.2 Heat capacities In 1819, science had a deceptive simplicity. Dulong and Petit, for example, were able to propose their law that ‘the atoms of all simple bodies have exactly the same heat capacity’ of about 25 J K1 mol1 (in modern units). Dulong and Petit’s rather primitive observations, though, were done at room temperature, and it was unfortunate for them and for classical physics when measurements were extended to lower temperatures and to a wider range of materials. It was found that all elements had heat capacities lower than predicted by Dulong and Petit’s law and that the values tended towards zero as T ! 0. Dulong and Petit’s law was easy to explain in terms of classical physics by assuming that each atom acts as a classical oscillator in three dimensions. The calculation predicted that the molar isochoric (constant volume) heat capacity, CV,m, of a monatomic solid should be equal to 3R ¼ 24.94 J K1 mol1, where R is the gas constant (R ¼ NAk, with NA Avogadro’s constant). That the heat capacities were smaller than predicted was a serious embarrassment. Einstein recognized the similarity between this problem and black-body radiation, for if each atomic oscillator required a certain minimum energy before it would actively oscillate and hence contribute to the heat capacity, then at low temperatures some would be inactive and the heat capacity would be smaller than expected. He applied Planck’s suggestion for electromagnetic oscillators to the material, atomic oscillators of the solid, and deduced the following expression:
3 Debye Einstein
CV,m /R
2
1
0
0
CV;m ðTÞ ¼ 3RfE ðTÞ
0.5
1 T /
1.5
Fig. 0.3 The Einstein and Debye molar heat capacities. The symbol y denotes the Einstein and Debye temperatures, respectively. Close to T ¼ 0 the Debye heat capacity is proportional to T3.
2
fE ðTÞ ¼
2 yE eyE =2T T 1 eyE =T
ð0:5aÞ
where the Einstein temperature, yE, is related to the frequency of atomic oscillators by yE ¼ hn/k. The function CV,m(T)/R is plotted in Fig. 0.3, and closely reproduces the experimental curve. In fact, the fit is not particularly good at very low temperatures, but that can be traced to Einstein’s assumption that all the atoms oscillated with the same frequency. When this restriction was removed by Debye, he obtained 3 Z yD =T T x4 ex dx CV;m ðTÞ ¼ 3RfD ðTÞ fD ðTÞ ¼ 3 x yD ðe 1Þ2 0
ð0:5bÞ
where the Debye temperature, yD, is related to the maximum frequency of the oscillations that can be supported by the solid. This expression gives a very good fit with observation. The importance of Einstein’s contribution is that it complemented Planck’s. Planck had shown that the energy of radiation is quantized;
4
j
INTRODUCTION AND ORIENTATION
Einstein showed that matter is quantized too. Quantization appears to be universal. Neither was able to justify the form that quantization took (with oscillators excitable in steps of hn), but that is a problem we shall solve later in the text.
0.3 The photoelectric and Compton effects In those enormously productive months of 1905–6, when Einstein formulated not only his theory of heat capacities but also the special theory of relativity, he found time to make another fundamental contribution to modern physics. His achievement was to relate Planck’s quantum hypothesis to the phenomenon of the photoelectric effect, the emission of electrons from metals when they are exposed to ultraviolet radiation. The puzzling features of the effect were that the emission was instantaneous when the radiation was applied however low its intensity, but there was no emission, whatever the intensity of the radiation, unless its frequency exceeded a threshold value typical of each element. It was also known that the kinetic energy of the ejected electrons varied linearly with the frequency of the incident radiation. Einstein pointed out that all the observations fell into place if the electromagnetic field was quantized, and that it consisted of bundles of energy of magnitude hn. These bundles were later named photons by G.N. Lewis, and we shall use that term from now on. Einstein viewed the photoelectric effect as the outcome of a collision between an incoming projectile, a photon of energy hn, and an electron buried in the metal. This picture accounts for the instantaneous character of the effect, because even one photon can participate in one collision. It also accounted for the frequency threshold, because a minimum energy (which is normally denoted F and called the ‘work function’ for the metal, the analogue of the ionization energy of an atom) must be supplied in a collision before photoejection can occur; hence, only radiation for which hn > F can be successful. The linear dependence of the kinetic energy, EK, of the photoelectron on the frequency of the radiation is a simple consequence of the conservation of energy, which implies that EK ¼ hn F
ð0:6Þ
If photons do have a particle-like character, then they should possess a linear momentum, p. The relativistic expression relating a particle’s energy to its mass and momentum is E2 ¼ m2 c4 þ p2 c2
ð0:7Þ
where c is the speed of light. In the case of a photon, E ¼ hn and m ¼ 0, so p¼
hn h ¼ c l
ð0:8Þ
0.4 ATOMIC SPECTRA
j
5
This linear momentum should be detectable if radiation falls on an electron, for a partial transfer of momentum during the collision should appear as a change in wavelength of the photons. In 1923, A.H. Compton performed the experiment with X-rays scattered from the electrons in a graphite target, and found the results fitted the following formula for the shift in wavelength, dl ¼ lf li, when the radiation was scattered through an angle y: dl ¼ 2lC sin2 12 y
ð0:9Þ
where lC ¼ h/mec is called the Compton wavelength of the electron (lC ¼ 2.426 pm). This formula is derived on the supposition that a photon does indeed have a linear momentum h/l and that the scattering event is like a collision between two particles. There seems little doubt, therefore, that electromagnetic radiation has properties that classically would have been characteristic of particles. The photon hypothesis seems to be a denial of the extensive accumulation of data that apparently provided unequivocal support for the view that electromagnetic radiation is wave-like. By following the implications of experiments and quantum concepts, we have accounted quantitatively for observations for which classical physics could not supply even a qualitative explanation.
0.4 Atomic spectra There was yet another body of data that classical physics could not elucidate before the introduction of quantum theory. This puzzle was the observation that the radiation emitted by atoms was not continuous but consisted of discrete frequencies, or spectral lines. The spectrum of atomic hydrogen had a very simple appearance, and by 1885 J. Balmer had already noticed that their wavenumbers, ~n, where ~n ¼ n/c, fitted the expression 1 1 ~n ¼ RH 2 2 ð0:10Þ 2 n where RH has come to be known as the Rydberg constant for hydrogen (RH ¼ 1.097 105 cm1) and n ¼ 3, 4, . . . . Rydberg’s name is commemorated because he generalized this expression to accommodate all the transitions in atomic hydrogen. Even more generally, the Ritz combination principle states that the frequency of any spectral line could be expressed as the difference between two quantities, or terms: ~n ¼ T1 T2
ð0:11Þ
This expression strongly suggests that the energy levels of atoms are confined to discrete values, because a transition from one term of energy hcT1 to another of energy hcT2 can be expected to release a photon of energy hc~n, or hn, equal to the difference in energy between the two terms: this argument
6
j
INTRODUCTION AND ORIENTATION
leads directly to the expression for the wavenumber of the spectroscopic transitions. But why should the energy of an atom be confined to discrete values? In classical physics, all energies are permissible. The first attempt to weld together Planck’s quantization hypothesis and a mechanical model of an atom was made by Niels Bohr in 1913. By arbitrarily assuming that the angular momentum of an electron around a central nucleus (the picture of an atom that had emerged from Rutherford’s experiments in 1910) was confined to certain values, he was able to deduce the following expression for the permitted energy levels of an electron in a hydrogen atom: En ¼
me4 1 8h2 e20 n2
n ¼ 1, 2, . . .
ð0:12Þ
where 1/m ¼ 1/me þ 1/mp and e0 is the vacuum permittivity, a fundamental constant. This formula marked the first appearance in quantum mechanics of a quantum number, n, which identifies the state of the system and is used to calculate its energy. Equation 0.12 is consistent with Balmer’s formula and accounted with high precision for all the transitions of hydrogen that were then known. Bohr’s achievement was the union of theories of radiation and models of mechanics. However, it was an arbitrary union, and we now know that it is conceptually untenable (for instance, it is based on the view that an electron travels in a circular path around the nucleus). Nevertheless, the fact that he was able to account quantitatively for the appearance of the spectrum of hydrogen indicated that quantum mechanics was central to any description of atomic phenomena and properties.
0.5 The duality of matter The grand synthesis of these ideas and the demonstration of the deep links that exist between electromagnetic radiation and matter began with Louis de Broglie, who proposed on the basis of relativistic considerations that with any moving body there is ‘associated a wave’, and that the momentum of the body and the wavelength are related by the de Broglie relation: l¼
h p
ð0:13Þ
We have seen this formula already (eqn 0.8), in connection with the properties of photons. De Broglie proposed that it is universally applicable. The significance of the de Broglie relation is that it summarizes a fusion of opposites: the momentum is a property of particles; the wavelength is a property of waves. This duality, the possession of properties that in classical physics are characteristic of both particles and waves, is a persistent theme in the interpretation of quantum mechanics. It is probably best to regard the terms ‘wave’ and ‘particle’ as remnants of a language based on a false
0.5 THE DUALITY OF MATTER
j
7
(classical) model of the universe, and the term ‘duality’ as a late attempt to bring the language into line with a current (quantum mechanical) model. The experimental results that confirmed de Broglie’s conjecture are the observation of the diffraction of electrons by the ranks of atoms in a metal crystal acting as a diffraction grating. Davisson and Germer, who performed this experiment in 1925 using a crystal of nickel, found that the diffraction pattern was consistent with the electrons having a wavelength given by the de Broglie relation. Shortly afterwards, G.P. Thomson also succeeded in demonstrating the diffraction of electrons by thin films of celluloid and gold.2 If electrons—if all particles—have wave-like character, then we should expect there to be observational consequences. In particular, just as a wave of definite wavelength cannot be localized at a point, we should not expect an electron in a state of definite linear momentum (and hence wavelength) to be localized at a single point. It was pursuit of this idea that led Werner Heisenberg to his celebrated uncertainty principle, that it is impossible to specify the location and linear momentum of a particle simultaneously with arbitrary precision. In other words, information about location is at the expense of information about momentum, and vice versa. This complementarity of certain pairs of observables, the mutual exclusion of the specification of one property by the specification of another, is also a major theme of quantum mechanics, and almost an icon of the difference between it and classical mechanics, in which the specification of exact trajectories was a central theme. The consummation of all this faltering progress came in 1926 when Werner Heisenberg and Erwin Schro¨dinger formulated their seemingly different but equally successful versions of quantum mechanics. These days, we step between the two formalisms as the fancy takes us, for they are mathematically equivalent, and each one has particular advantages in different types of calculation. Although Heisenberg’s formulation preceded Schro¨dinger’s by a few months, it seemed more abstract and was expressed in the then unfamiliar vocabulary of matrices. Still today it is more suited for the more formal manipulations and deductions of the theory, and in the following pages we shall employ it in that manner. Schro¨dinger’s formulation, which was in terms of functions and differential equations, was more familiar in style but still equally revolutionary in implication. It is more suited to elementary manipulations and to the calculation of numerical results, and we shall employ it in that manner. ‘Experiments’, said Planck, ‘are the only means of knowledge at our disposal. The rest is poetry, imagination.’ It is time for that imagination to unfold.
.......................................................................................................
2. It has been pointed out by M. Jammer that J.J. Thomson was awarded the Nobel Prize for showing that the electron is a particle, and G.P. Thomson, his son, was awarded the Prize for showing that the electron is a wave. (See The conceptual development of quantum mechanics, McGraw-Hill, New York (1966), p. 254.)
8
j
INTRODUCTION AND ORIENTATION
PROBLEMS 0.1 Calculate the size of the quanta involved in the excitation of (a) an electronic motion of period 1.0 fs, (b) a molecular vibration of period 10 fs, and (c) a pendulum of period 1.0 s. 0.2 Find the wavelength corresponding to the maximum in the Planck distribution for a given temperature, and show that the expression reduces to the Wien displacement law at short wavelengths. Determine an expression for the constant in the law in terms of fundamental constants. (This constant is called the second radiation constant, c2.) 0.3 Use the Planck distribution to confirm the Stefan–Boltzmann law and to derive an expression for the Stefan–Boltzmann constant s. 0.4 The peak in the Sun’s emitted energy occurs at about 480 nm. Estimate the temperature of its surface on the basis of it being regarded as a black-body emitter. 0.5 Derive the Einstein formula for the heat capacity of a collection of harmonic oscillators. To do so, use the quantum mechanical result that the energy of a harmonic oscillator of force constant k and mass m is one of the values (v þ 12)hv, with v ¼ (1/2p)(k/m)1/2 and v ¼ 0, 1, 2, . . . . Hint. Calculate the mean energy, E, of a collection of oscillators by substituting these energies into the Boltzmann distribution, and then evaluate C ¼ dE/dT. 0.6 Find the (a) low temperature, (b) high temperature forms of the Einstein heat capacity function. 0.7 Show that the Debye expression for the heat capacity is proportional to T3 as T ! 0. 0.8 Estimate the molar heat capacities of metallic sodium (yD ¼ 150 K) and diamond (yD ¼ 1860 K) at room temperature (300 K). 0.9 Calculate the molar entropy Rof an Einstein solid at T T ¼ yE. Hint. The entropy is S ¼ 0 ðCV =TÞdT. Evaluate the integral numerically. 0.10 How many photons would be emitted per second by a sodium lamp rated at 100 W which radiated all its energy with 100 per cent efficiency as yellow light of wavelength 589 nm? 0.11 Calculate the speed of an electron emitted from a clean potassium surface (F ¼ 2.3 eV) by light of wavelength (a) 300 nm, (b) 600 nm. 0.12 When light of wavelength 195 nm strikes a certain metal surface, electrons are ejected with a speed of 1.23 106 m s1. Calculate the speed of electrons ejected from the same metal surface by light of wavelength 255 nm.
0.13 At what wavelength of incident radiation do the relativistic and non-relativistic expressions for the ejection of electrons from potassium differ by 10 per cent? That is, find l such that the non-relativistic and relativistic linear momenta of the photoelectron differ by 10 per cent. Use F ¼ 2.3 eV. 0.14 Deduce eqn 0.9 for the Compton effect on the basis of the conservation of energy and linear momentum. Hint. Use the relativistic expressions. Initially the electron is at rest with energy mec2. When it is travelling with momentum p its energy is ðp2 c2 þ m2e c4 Þ1/2. The photon, with initial momentum h/li and energy hni, strikes the stationary electron, is deflected through an angle y, and emerges with momentum h/lf and energy hnf. The electron is initially stationary (p ¼ 0) but moves off with an angle y 0 to the incident photon. Conserve energy and both components of linear momentum. Eliminate y 0 , then p, and so arrive at an expression for dl. 0.15 The first few lines of the visible (Balmer) series in the spectrum of atomic hydrogen lie at l/nm ¼ 656.46, 486.27, 434.17, 410.29, . . . . Find a value of RH, the Rydberg constant for hydrogen. The ionization energy, I, is the minimum energy required to remove the electron. Find it from the data and express its value in electron volts. How is I related to RH? Hint. The ionization limit corresponds to n ! 1 for the final state of the electron. 0.16 Calculate the de Broglie wavelength of (a) a mass of 1.0 g travelling at 1.0 cm s1, (b) the same at 95 per cent of the speed of light, (c) a hydrogen atom at room temperature (300 K); estimate the mean speed from the equipartition principle, which implies that the mean kinetic energy of an atom is equal to 32kT, where k is Boltzmann’s constant, (d) an electron accelerated from rest through a potential difference of (i) 1.0 V, (ii) 10 kV. Hint. For the momentum in (b) use p ¼ mv/(l v2/c2)1/2 and for the speed in (d) use 2 1 2mev ¼ eV, where V is the potential difference. 0.17 Derive eqn 0.12 for the permitted energy levels for the electron in a hydrogen atom. To do so, use the following (incorrect) postulates of Bohr: (a) the electron moves in a circular orbit of radius r around the nucleus and (b) the angular momentum of the electron is an integral multiple of h, that is me vr ¼ n h. Hint. Mechanical stability of the orbital motion requires that the Coulombic force of attraction between the electron and nucleus equals the centrifugal force due to the circular motion. The energy of the electron is the sum of the kinetic energy and potential (Coulombic) energy. For simplicity, use me rather than the reduced mass m.
1 Operators in quantum mechanics 1.1 Linear operators 1.2 Eigenfunctions and eigenvalues 1.3 Representations 1.4 Commutation and non-commutation 1.5 The construction of operators 1.6 Integrals over operators 1.7 Dirac bracket notation 1.8 Hermitian operators The postulates of quantum mechanics 1.9 States and wavefunctions 1.10 The fundamental prescription 1.11 The outcome of measurements 1.12 The interpretation of the wavefunction 1.13 The equation for the wavefunction 1.14 The separation of the Schro¨dinger equation The specification and evolution of states 1.15 Simultaneous observables 1.16 The uncertainty principle 1.17 Consequences of the uncertainty principle 1.18 The uncertainty in energy and time 1.19 Time-evolution and conservation laws Matrices in quantum mechanics 1.20 Matrix elements 1.21 The diagonalization of the hamiltonian The plausibility of the Schro¨dinger equation 1.22 The propagation of light 1.23 The propagation of particles 1.24 The transition to quantum mechanics
The foundations of quantum mechanics
The whole of quantum mechanics can be expressed in terms of a small set of postulates. When their consequences are developed, they embrace the behaviour of all known forms of matter, including the molecules, atoms, and electrons that will be at the centre of our attention in this book. This chapter introduces the postulates and illustrates how they are used. The remaining chapters build on them, and show how to apply them to problems of chemical interest, such as atomic and molecular structure and the properties of molecules. We assume that you have already met the concepts of ‘hamiltonian’ and ‘wavefunction’ in an elementary introduction, and have seen the Schro¨dinger equation written in the form Hc ¼ Ec This chapter establishes the full significance of this equation, and provides a foundation for its application in the following chapters.
Operators in quantum mechanics An observable is any dynamical variable that can be measured. The principal mathematical difference between classical mechanics and quantum mechanics is that whereas in the former physical observables are represented by functions (such as position as a function of time), in quantum mechanics they are represented by mathematical operators. An operator is a symbol for an instruction to carry out some action, an operation, on a function. In most of the examples we shall meet, the action will be nothing more complicated than multiplication or differentiation. Thus, one typical operation might be multiplication by x, which is represented by the operator x . Another operation might be differentiation with respect to x, represented by the operator d/dx. We shall represent operators by the symbol O (omega) in general, but use A, B, . . . when we want to refer to a series of operators. We shall not in general distinguish between the observable and the operator that represents that observable; so the position of a particle along the x-axis will be denoted x and the corresponding operator will also be denoted x (with multiplication implied). We shall always make it clear whether we are referring to the observable or the operator. We shall need a number of concepts related to operators and functions on which they operate, and this first section introduces some of the more important features.
10
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
1.1 Linear operators The operators we shall meet in quantum mechanics are all linear. A linear operator is one for which Oðaf þ bgÞ ¼ aOf þ bOg
ð1:1Þ
where a and b are constants and f and g are functions. Multiplication is a linear operation; so is differentiation and integration. An example of a nonlinear operation is that of taking the logarithm of a function, because it is not true, for example, that log 2x ¼ 2 log x for all x.
1.2 Eigenfunctions and eigenvalues In general, when an operator operates on a function, the outcome is another function. Differentiation of sin x, for instance, gives cos x. However, in certain cases, the outcome of an operation is the same function multiplied by a constant. Functions of this kind are called ‘eigenfunctions’ of the operator. More formally, a function f (which may be complex) is an eigenfunction of an operator O if it satisfies an equation of the form Of ¼ of
ð1:2Þ
where o is a constant. Such an equation is called an eigenvalue equation. The function eax is an eigenfunction of the operator d/dx because (d/dx)eax ¼ aeax, 2 which is a constant (a) multiplying the original function. In contrast, eax is ax2 ax2 not an eigenfunction of d/dx, because (d/dx)e ¼ 2axe , which is a con2 stant (2a) times a different function of x (the function xeax ). The constant o in an eigenvalue equation is called the eigenvalue of the operator O. Example 1.1 Determining if a function is an eigenfunction
Is the function cos(3x þ 5) an eigenfunction of the operator d2/dx2 and, if so, what is the corresponding eigenvalue? Method. Perform the indicated operation on the given function and see if
the function satisfies an eigenvalue equation. Use (d/dx)sin ax ¼ a cos ax and (d/dx)cos ax ¼ a sin ax. Answer. The operator operating on the function yields
d2 d ð3 sinð3x þ 5ÞÞ ¼ 9 cosð3x þ 5Þ cosð3x þ 5Þ ¼ 2 dx dx and we see that the original function reappears multiplied by the eigenvalue 9. Self-test 1.1. Is the function e3x þ 5 an eigenfunction of the operator d2/dx2
and, if so, what is the corresponding eigenvalue? [Yes; 9]
An important point is that a general function can be expanded in terms of all the eigenfunctions of an operator, a so-called complete set of functions.
1.2 EIGENFUNCTIONS AND EIGENVALUES
j
11
That is, if fn is an eigenfunction of an operator O with eigenvalue on (so Ofn ¼ on fn), then1 a general function g can be expressed as the linear combination X g¼ cn fn ð1:3Þ n
where the cn are coefficients and the sum is over a complete set of functions. For instance, the straight line g ¼ ax can be recreated over a certain range by superimposing an infinite number of sine functions, each of which is an eigenfunction of the operator d2/dx2. Alternatively, the same function may be constructed from an infinite number of exponential functions, which are eigenfunctions of d/dx. The advantage of expressing a general function as a linear combination of a set of eigenfunctions is that it allows us to deduce the effect of an operator on a function that is not one of its own eigenfunctions. Thus, the effect of O on g in eqn 1.3, using the property of linearity, is simply X X X cn fn ¼ cn Ofn ¼ c n on f n Og ¼ O n
n
n
A special case of these linear combinations is when we have a set of degenerate eigenfunctions, a set of functions with the same eigenvalue. Thus, suppose that f1, f2, . . . , fk are all eigenfunctions of the operator O, and that they all correspond to the same eigenvalue o: Ofn ¼ ofn with n ¼ 1, 2, . . . , k
ð1:4Þ
Then it is quite easy to show that any linear combination of the functions fn is also an eigenfunction of O with the same eigenvalue o. The proof is as follows. For an arbitrary linear combination g of the degenerate set of functions, we can write Og ¼ O
k X n¼1
cn fn ¼
k X n¼1
cn Ofn ¼
k X
cn ofn ¼ o
n¼1
k X
cn fn ¼ og
n¼1
This expression has the form of an eigenvalue equation (Og ¼ og). Example 1.2 Demonstrating that a linear combination of degenerate
eigenfunctions is also an eigenfunction Show that any linear combination of the complex functions e2ix and e2ix is an eigenfunction of the operator d2/dx2, where i ¼ (1)1/2. Method. Consider an arbitrary linear combination ae2ix þ be2ix and see if the
function satisfies an eigenvalue equation. Answer. First we demonstrate that e2ix and e2ix are degenerate eigenfunctions.
d2 2ix d ð2ie2ix Þ ¼ 4e2ix e ¼ dx dx2
.......................................................................................................
1. See P.M. Morse and H. Feschbach, Methods of theoretical physics, McGraw-Hill, New York (1953).
12
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
where we have used i2 ¼ 1. Both functions correspond to the same eigenvalue, 4. Then we operate on a linear combination of the functions. d2 ðae2ix þ be2ix Þ ¼ 4ðae2ix þ be2ix Þ dx2 The linear combination satisfies the eigenvalue equation and has the same eigenvalue (4) as do the two complex functions. Self-test 1.2. Show that any linear combination of the functions sin(3x) and
cos(3x) is an eigenfunction of the operator d2/dx2. [Eigenvalue is 9]
A further technical point is that from n basis functions it is possible to construct n linearly independent combinations. A set of functions g1, g2, . . . , gn is said to be linearly independent if we cannot find a set of constants c1, c2, . . . , cn (other than the trivial set c1 ¼ c2 ¼ ¼ 0) for which X ci gi ¼ 0 i
A set of functions that is not linearly independent is said to be linearly dependent. From a set of n linearly independent functions, it is possible to construct an infinite number of sets of linearly independent combinations, but each set can have no more than n members. For example, from three 2p-orbitals of an atom it is possible to form any number of sets of linearly independent combinations, but each set has no more than three members.
1.3 Representations The remaining work of this section is to put forward some explicit forms of the operators we shall meet. Much of quantum mechanics can be developed in terms of an abstract set of operators, as we shall see later. However, it is often fruitful to adopt an explicit form for particular operators and to express them in terms of the mathematical operations of multiplication, differentiation, and so on. Different choices of the operators that correspond to a particular observable give rise to the different representations of quantum mechanics, because the explicit forms of the operators represent the abstract structure of the theory in terms of actual manipulations. One of the most common representations is the position representation, in which the position operator is represented by multiplication by x (or whatever coordinate is specified) and the linear momentum parallel to x is represented by differentiation with respect to x. Explicitly: q h ð1:5Þ i qx where h ¼ h=2p. Why the linear momentum should be represented in precisely this manner will be explained in the following section. For the time being, it may be taken to be a basic postulate of quantum mechanics. An alternative choice of operators is the momentum representation, in which the linear momentum parallel to x is represented by the operation of Position representation: x ! x
px !
1.4 COMMUTATION AND NON-COMMUTATION
j
13
multiplication by px and the position operator is represented by differentiation with respect to px. Explicitly: Momentum representation: x !
q h i qpx
px ! px
ð1:6Þ
There are other representations. We shall normally use the position representation when the adoption of a representation is appropriate, but we shall also see that many of the calculations in quantum mechanics can be done independently of a representation.
1.4 Commutation and non-commutation An important feature of operators is that in general the outcome of successive operations (A followed by B, which is denoted BA, or B followed by A, denoted AB) depends on the order in which the operations are carried out. That is, in general BA 6¼ AB. We say that, in general, operators do not commute. For example, consider the operators x and px and a specific h/i)x ¼ (2 h/i)x2, function x2. In the position representation, (xpx)x2 ¼ x(2 2 3 2 h/i)x . The operators x and px do not commute. whereas (pxx)x ¼ pxx ¼ (3 The quantity AB BA is called the commutator of A and B and is denoted [A, B]: ½A, B ¼ AB BA
ð1:7Þ
It is instructive to evaluate the commutator of the position and linear momentum operators in the two representations shown above; the procedure is illustrated in the following example. Example 1.3 The evaluation of a commutator
Evaluate the commutator [x,px] in the position representation. Method. To evaluate the commutator [A,B] we need to remember that the
operators operate on some function, which we shall write f. So, evaluate [A,B]f for an arbitrary function f, and then cancel f at the end of the calculation. Answer. Substitution of the explicit expressions for the operators into [x,px] proceeds as follows: h qf h qðxf Þ ½x, px f ¼ ðxpx px xÞf ¼ x i qx i qx h qf h h qf f x ¼ i hf ¼x i qx i i qx
where we have used (1/i) ¼ i. This derivation is true for any function f, so in terms of the operators themselves, ½x, px ¼ ih The right-hand side should be interpreted as the operator ‘multiply by the constant ih’. Self-test 1.3. Evaluate the same commutator in the momentum representation. [Same]
14
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
1.5 The construction of operators Operators for other observables of interest can be constructed from the operators for position and momentum. For example, the kinetic energy operator T can be constructed by noting that kinetic energy is related to linear momentum by T ¼ p2/2m where m is the mass of the particle. It follows that in one dimension and in the position representation p2 1 h d 2 h d2 ¼ ð1:8Þ T¼ x ¼ 2m dx2 2m 2m i dx
Although eqn 1.9 has explicitly used Cartesian coordinates, the relation between the kinetic energy operator and the laplacian is true in any coordinate system; for example, spherical polar coordinates.
In three dimensions the operator in the position representation is ( ) h 2 q2 q2 q2 h2 2 r T¼ þ 2þ 2 ¼ 2 2m qx qy qz 2m
ð1:9Þ
The operator r2, which is read ‘del squared’ and called the laplacian, is the sum of the three second derivatives. The operator for potential energy of a particle in one dimension, V(x), is multiplication by the function V(x) in the position representation. The same is true of the potential energy operator in three dimensions. For example, in the position representation the operator for the Coulomb potential energy of an electron (charge e) in the field of a nucleus of atomic number Z is the multiplicative operator V¼
Ze2 4pe0 r
ð1:10Þ
where r is the distance from the nucleus to the electron. It is usual to omit the multiplication sign from multiplicative operators, but it should not be forgotten that such expressions are multiplications. The operator for the total energy of a system is called the hamiltonian operator and is denoted H: H ¼TþV
ð1:11Þ
The name commemorates W.R. Hamilton’s contribution to the formulation of classical mechanics in terms of what became known as a hamiltonian function. To write the explicit form of this operator we simply substitute the appropriate expressions for the kinetic and potential energy operators in the chosen representation. For example, the hamiltonian for a particle of mass m moving in one dimension is H¼
2 d2 h þ VðxÞ 2m dx2
ð1:12Þ
where V(x) is the operator for the potential energy. Similarly, the hamiltonian operator for an electron of mass me in a hydrogen atom is H¼
h 2 2 e2 r 2me 4pe0 r
ð1:13Þ
1.6 INTEGRALS OVER OPERATORS
j
15
The general prescription for constructing operators in the position representation should be clear from these examples. In short: 1. Write the classical expression for the observable in terms of position coordinates and the linear momentum. h/i)q/qx (and likewise 2. Replace x by multiplication by x, and replace px by ( for the other coordinates).
1.6 Integrals over operators When we want to make contact between a calculation done using operators and the actual outcome of an experiment, we need to evaluate certain integrals. These integrals all have the form Z ð1:14Þ I ¼ fm Ofn dt The complex conjugate of a complex number z ¼ a þ ib is z ¼ a ib. Complex conjugation amounts to everywhere replacing i by i. The square modulus jzj2 is given by zz ¼ a2 þ b2 since jij2 ¼ 1.
where fm is the complex conjugate of fm. In this integral dt is the volume element. In one dimension, dt can be identified as dx; in three dimensions it is dxdydz. The integral is taken over the entire space available to the system, which is typically from x ¼ 1 to x ¼ þ 1 (and similarly for the other coordinates). A glance at the later pages of this book will show that many molecular properties are expressed as combinations of integrals of this form (often in a notation which will be explained later). Certain special cases of this type of integral have special names, and we shall introduce them here. When the operator O in eqn 1.14 is simply multiplication by 1, the integral is called an overlap integral and commonly denoted S: Z ð1:15Þ S ¼ fm fn dt It is helpful to regard S as a measure of the similarity of two functions: when S ¼ 0, the functions are classified as orthogonal, rather like two perpendicular vectors. When S is close to 1, the two functions are almost identical. The recognition of mutually orthogonal functions often helps to reduce the amount of calculation considerably, and rules will emerge in later sections and chapters. The normalization integral is the special case of eqn 1.15 for m ¼ n. A function fm is said to be normalized (strictly, normalized to 1) if Z fm fm dt ¼ 1 ð1:16Þ It is almost always easy to ensure that a function is normalized by multiplying it by an appropriate numerical factor, which is called a normalization factor, typically denoted N and taken to be real so that N ¼ N. The procedure is illustrated in the following example. Example 1.4 How to normalize a function
A certain function f is sin(px/L) between x ¼ 0 and x ¼ L and is zero elsewhere. Find the normalized form of the function.
16
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
Method. We need to find the (real) factor N such that N sin(px/L) is normalized to 1. To find N we substitute this expression into eqn 1.16, evaluate the integral, and select N to ensure normalization. Note that ‘all space’ extends from x ¼ 0 to x ¼ L. Answer. The necessary integration is
Z
Z
L
N 2 sin2 ðpx=LÞdx ¼ 12LN 2 R where we have used sin2ax dx ¼ (x/2)(sin 2ax)/4a þ constant. For this integral to be equal to 1, we require N ¼ (2/L)1/2. The normalized function is therefore 1=2 2 f ¼ sinðpx=LÞ L f f dt ¼
0
Comment. We shall see later that this function describes the distribution of a
particle in a square well, and we shall need its normalized form there. Self-test 1.4. Normalize the function f ¼ eif, where f ranges from 0 to 2p. [N ¼ 1/(2p)1/2]
A set of functions fn that are (a) normalized and (b) mutually orthogonal are said to satisfy the orthonormality condition: Z fm fn dt ¼ dmn ð1:17Þ In this expression, dmn denotes the Kronecker delta, which is 1 when m ¼ n and 0 otherwise.
1.7 Dirac bracket notation With eqn 1.14 we are on the edge of getting lost in a complicated notation. The appearance of many quantum mechanical expressions is greatly simplified by adopting the Dirac bracket notation in which integrals are written as follows: Z hmjOjni ¼ fm Ofn dt ð1:18Þ The symbol jni is called a ket, and denotes the state described by the function fn. Similarly, the symbol hnj is called a bra, and denotes the complex conjugate of the function, fn . When a bra and ket are strung together with an operator between them, as in the bracket hmjOjni, the integral in eqn 1.18 is to be understood. When the operator is simply multiplication by 1, the 1 is omitted and we use the convention Z ð1:19Þ hmjni ¼ fm fn dt This notation is very elegant. For example, the normalization integral becomes hnjni ¼ 1 and the orthogonality condition becomes hmjni ¼ 0 for m 6¼ n. The combined orthonormality condition (eqn 1.17) is then hmjni ¼ dmn
ð1:20Þ
1.8 HERMITIAN OPERATORS
j
17
A final point is that, as can readily be deduced from the definition of a Dirac bracket, hmjni ¼ hnjmi
1.8 Hermitian operators An operator is hermitian if it satisfies the following relation: Z Z fm Ofn dt ¼ fn Ofm dt
ð1:21aÞ
for any two functions fm and fn. An alternative version of this definition is Z Z ð1:21bÞ fm Ofn dt ¼ ðOfm Þ fn dt This expression is obtained by taking the complex conjugate of each term on the right-hand side of eqn 1.21a. In terms of the Dirac notation, the definition of hermiticity is hmjOjni ¼ hnjOjmi
ð1:22Þ
Example 1.5 How to confirm the hermiticity of operators
Show that the position and momentum operators in the position representation are hermitian. Method. We need to show that the operators satisfy eqn 1.21a. In some cases
(the position operator, for instance), the hermiticity is obvious as soon as the integral is written down. When a differential operator is used, it may be necessary to use integration by parts at some stage in the argument to transfer the differentiation from one function to another: Z Z u dv ¼ uv v du Answer. That the position operator is hermitian is obvious from inspection:
Z
fm xfn dt ¼
Z
fn xfm dt ¼
Z
fn xfm dt
We have used the facts that (f ) ¼ f and x is real. The demonstration of the hermiticity of px, a differential operator in the position representation, involves an integration by parts: Z
Z
Z d h h fn dx ¼ fm dfn i dx i x¼1 Z
h
¼ f fn fn dfm
i m x¼1 Z 1 h x¼1 d ¼ fn fm dx fm fn jx¼1 i dx 1
fm px fn dx ¼
fm
18
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
The first term on the right is zero (because when jxj is infinite, a normalizable function must be vanishingly small; see Section 1.12). Therefore, Z
h i Z
fm px fn dx ¼ ¼
Z
d f dx dx m Z h d fn px fm dx fm dx ¼ fn i dx fn
Hence, the operator is hermitian. Self-test 1.5. Show that the two operators are hermitian in the momentum
representation.
As we shall now see, the property of hermiticity has far-reaching implications. First, we shall establish the following property: Property 1. The eigenvalues of hermitian operators are real. Proof 1.1 The reality of eigenvalues
Consider the eigenvalue equation Ojoi ¼ ojoi The ket joi denotes an eigenstate of the operator O in the sense that the corresponding function fo is an eigenfunction of the operator O and we are labelling the eigenstates with the eigenvalue o of the operator O. It is often convenient to use the eigenvalues as labels in this way. Multiplication from the left by hoj results in the equation hojOjoi ¼ ohojoi ¼ o taking joi to be normalized. Now take the complex conjugate of both sides: hojOjoi ¼ o However, by hermiticity, hojOjoi ¼ hojOjoi. Therefore, it follows that o ¼ o , which implies that the eigenvalue o is real.
The second property we shall prove is as follows: Property 2. Eigenfunctions corresponding to different eigenvalues of an hermitian operator are orthogonal. That is, if we have two eigenfunctions of an hermitian operator O with eigenvalues o and o 0 , with o 6¼ o 0 , then hojo 0 i ¼ 0. For example, it follows at once that all the eigenfunctions of a harmonic oscillator (Section 2.16) are mutually orthogonal, for as we shall see each one corresponds to a different energy (the eigenvalue of the hamiltonian, an hermitian operator).
1.9 STATES AND WAVEFUNCTIONS
j
19
Proof 1.2 The orthogonality of eigenstates
Suppose we have two eigenstates joi and jo 0 i that satisfy the following relations: Ojoi ¼ ojoi and
Ojo0 i ¼ o0 jo0 i
Then multiplication of the first relation by ho 0 j and the second by hoj gives ho0 jOjoi ¼ oho0 joi and
hojOjo0 i ¼ o0 hojo0 i
Now take the complex conjugate of the second relation and subtract it from the first while using Property 1 (o 0 ¼ o 0 ): ho0 jOjoi hojOjo0 i ¼ oho0 joi o0 hojo0 i Because O is hermitian, the left-hand side of this expression is zero; so (noting that o 0 is real and using hojo 0 i ¼ ho 0 joi as explained earlier) we arrive at ðo o0 Þho0 joi ¼ 0 However, because the two eigenvalues are different, the only way of satisfying this relation is for ho 0 joi ¼ 0, as was to be proved.
The postulates of quantum mechanics Now we turn to an application of the preceding material, and move into the foundations of quantum mechanics. The postulates we use as a basis for quantum mechanics are by no means the most subtle that have been devised, but they are strong enough for what we have to do.
1.9 States and wavefunctions The first postulate concerns the information we can know about a state: Postulate 1. The state of a system is fully described by a function C(r1, r2, . . . , t). In this statement, r1, r2, . . . are the spatial coordinates of particles 1, 2, . . . that constitute the system and t is the time. The function C (uppercase psi) plays a central role in quantum mechanics, and is called the wavefunction of the system (more specifically, the time-dependent wavefunction). When we are not interested in how the system changes in time we shall denote the wavefunction by a lowercase psi as c(r1, r2, . . . ) and refer to it as the timeindependent wavefunction. The state of the system may also depend on some internal variable of the particles (their spin states); we ignore that for now and return to it later. By ‘describe’ we mean that the wavefunction contains information about all the properties of the system that are open to experimental determination. We shall see that the wavefunction of a system will be specified by a set of labels called quantum numbers, and may then be written ca,b, . . . , where a, b, . . . are the quantum numbers. The values of these quantum numbers specify the wavefunction and thus allow the values of various physical
20
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
observables to be calculated. It is often convenient to refer to the state of the system without referring to the corresponding wavefunction; the state is specified by listing the values of the quantum numbers that define it.
1.10 The fundamental prescription The next postulate concerns the selection of operators: Postulate 2. Observables are represented by hermitian operators chosen to satisfy the commutation relations ½q, pq0 ¼ ihdqq0
½q, q0 ¼ 0 ½pq , pq0 ¼ 0
where q and q 0 each denote one of the coordinates x, y, z and pq and pq 0 the corresponding linear momenta. The requirement that the operators are hermitian ensures that the observables have real values (see below). Each commutation relation is a basic, unprovable, and underivable postulate. Postulate 2 is the basis of the selection of the form of the operators in the position and momentum representations for all observables that depend on the position and the momentum.2 Thus, if we define the position representation as the representation in which the position operator is multiplication by the position coordinate, then as we saw in Example 1.3, it follows that the momentum operator must involve differentiation with respect to x, as specified earlier. Similarly, if the momentum representation is defined as the representation in which the linear momentum is represented by multiplication, then the form of the position operator is fixed as a derivative with respect to the linear momentum. The coordinates x, y, and z commute with each other as do the linear momenta px, py, and pz.
1.11 The outcome of measurements The next postulate brings together the wavefunction and the operators and establishes the link between formal calculations and experimental observations: Postulate 3. When a system is described by a wavefunction c, the mean value of the observable O in a series of measurements is equal to the expectation value of the corresponding operator. The expectation value of an operator O for an arbitrary state c is denoted hOi and defined as R c Oc dt hcjOjci hOi ¼ R ð1:23Þ ¼ hcjci c cdt If the wavefunction is chosen to be normalized to 1, then the expectation value is simply Z ð1:24Þ hOi ¼ c Oc dt ¼ hcjOjci Unless we state otherwise, from now on we shall assume that the wavefunction is normalized to 1. .......................................................................................................
2. This prescription excludes intrinsic observables, such as spin (Section 4.8).
1.11 THE OUTCOME OF MEASUREMENTS
j
21
The meaning of Postulate 3 can be unravelled as follows. First, suppose that c is an eigenfunction of O with eigenvalue o; then Z Z Z ð1:25Þ hOi ¼ c Oc dt ¼ c oc dt ¼ o c c dt ¼ o That is, a series of experiments on identical systems to determine O will give the average value o (a real quantity, because O is hermitian). Now suppose that although the system is in an eigenstate of the hamiltonian it is not in an eigenstate of O. In this case the wavefunction can be expressed as a linear combination of eigenfunctions of O: X cn cn where Ocn ¼ on cn c¼ n
In this case, the expectation value is ! ! Z Z X X X hOi ¼ cm cm O cn cn dt ¼ cm cn cm Ocn dt m n m, n Z X ¼ cm cn on cm cn dt m, n Because the eigenfunctions form an orthonormal set, the integral in the last expression is zero if n 6¼ m, is 1 if n ¼ m, and the double sum reduces to a single sum: Z X X 2 X cn cn on cn cn dt ¼ cn cn on ¼ jcn j on ð1:26Þ hOi ¼ n
n
n
That is, the expectation value is a weighted sum of the eigenvalues of O, the contribution of a particular eigenvalue to the sum being determined by the square modulus of the corresponding coefficient in the expansion of the wavefunction. We can now interpret the difference between eqns 1.25 and 1.26 in the form of a subsidiary postulate: Postulate 30 . When c is an eigenfunction of the operator O, the determination of the property O always yields one result, namely the corresponding eigenvalue o. The expectation value will simply be the eigenvalue o. When c is not an eigenfunction of O, a single measurement of the property yields a single outcome which is one of the eigenvalues of O, and the probability that a particular eigenvalue on is measured is equal to jcnj2, where cn is the coefficient of the eigenfunction cn in the expansion of the wavefunction. One measurement can give only one result: a pointer can indicate only one value on a dial at any instant. A series of determinations can lead to a series of results with some mean value. The subsidiary postulate asserts that a measurement of the observable O always results in the pointer indicating one of the eigenvalues of the corresponding operator. If the function that describes the state of the system is an eigenfunction of O, then every pointer reading is precisely o and the mean value is also o. If the system has been prepared in a state that is not an eigenfunction of O, then different measurements give different values, but every individual measurement is one of the eigenvalues of
22
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
O, and the probability that a particular outcome on is obtained is determined by the value of jcnj2. In this case, the mean value of all the observations is the weighted average of the eigenvalues. Note that in either case, the hermiticity of the operator guarantees that the observables will be real. Example 1.6 How to use Postulate 3 0 .
An operator A has eigenfunctions f1, f2, . . . , fn with corresponding eigenvalues a1, a2, . . . , an. The state of a system is described by a normalized wavefunction c given by
1=2
1=2 c ¼ 12 f1 38 f2 þ 38 i f3 What will be the outcome of measuring the observable A? Method. First, we need to determine if c is an eigenfunction of the operator A.
If it is, then we shall obtain the same eigenvalue of A in every measurement. If it is not, we shall obtain different values in a series of different measurements. In the latter case, if we have an expression for c in terms of the eigenfunctions of A, then we can determine what different values are possible, the probabilities of obtaining them, and the average value from a large series of measurements. Answer. To test whether c is an eigenfunction of the operator A we proceed as
follows: h
1=2
1=2 i Ac ¼ A 12 f1 38 f2 þ 38 i f3
1=2 1=2 a2 f2 þ 38 i a3 f3 6¼ constant c ¼ 12a1 f1 38 Therefore, c is not an eigenfunction of A. However, because c is a linear combination of f1, f2, and f3 we will obtain, in different measurements, the values a1, a2, and a3 (the eigenvalues of the eigenfunctions of A that contribute to c). The probabilities of obtaining a1, a2, and a3 are, respectively, 1 3 3 4, 8, and 8. The average value, given by eqn 1.26, is hAi ¼ 14 a1 þ 38 a2 þ 38 a3 Comment. The normalization of c is reflected in the fact that the probabilities
sum to 1. Because the eigenfunctions f4, f5, . . . do not contribute here to c, there is zero probability of finding a4, a5, . . . . Self-test 1.6. Repeat the problem using c ¼ 13 f2 þ ð79Þ
1=2
f4 13 if5 : [hAi ¼ 19 a2 þ 79 a4 þ 19 a5
1.12 The interpretation of the wavefunction The next postulate concerns the interpretation of the wavefunction itself, and is commonly called the Born interpretation: Postulate 4. The probability that a particle will be found in the volume element dt at the point r is proportional to jc(r)j2dt.
1.14 THE SEPARATION OF THE SCHRO¨ DINGER EQUATION
j
23
As we have already remarked, in one dimension the volume element is dx. In three dimensions the volume element is dxdydz. It follows from this interpretation that jc(r)j2 is a probability density, in the sense that it yields a probability when multiplied by the volume dt of an infinitesimal region. The wavefunction itself is a probability amplitude, and has no direct physical meaning. Note that whereas the probability density is real and nonnegative, the wavefunction may be complex and negative. It is usually convenient to use a normalized wavefunction; then the Born interpretation becomes an equality rather than a proportionality. The implication of the Born interpretation is that the wavefunction should be square-integrable; that is Z jcj2 dt < 1 because there must be a finite probability of finding the particle somewhere in the whole of space (and that probability is 1 for a normalized wavefunction). This postulate in turn implies that c ! 0 as x ! 1, for otherwise the integral of jcj2 would be infinite. We shall make frequent use of this implication throughout the text.
1.13 The equation for the wavefunction The final postulate concerns the dynamical evolution of the wavefunction with time: Postulate 5. The wavefunction C(r1, r2, . . . , t) evolves in time according to the equation ih
qC ¼ HC qt
ð1:27Þ
This partial differential equation is the celebrated Schro¨dinger equation introduced by Erwin Schro¨dinger in 1926. At this stage, we are treating the equation as an unmotivated postulate. However, in Section 1.24 we shall advance arguments in support of its plausibility. The operator H in the Schro¨dinger equation is the hamiltonian operator for the system, the operator corresponding to the total energy. For example, by using the expression in eqn 1.12, we obtain the time-dependent Schro¨dinger equation in one dimension (x) with a time-independent potential energy for a single particle: ih
qC h2 q2 C þ VðxÞC ¼ qt 2m qx2
ð1:28Þ
We shall have a great deal to say about the Schro¨dinger equation and its solutions in the rest of the text.
1.14 The separation of the Schro¨dinger equation The Schro¨dinger equation can often be separated into equations for the time and space variation of the wavefunction. The separation is possible when the potential energy is independent of time.
24
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
In one dimension the equation has the form HC ¼
2 q2 C h qC þ VðxÞC ¼ ih qt 2m qx2
Equations of this form can be solved by the technique of separation of variables, in which a trial solution takes the form Cðx, tÞ ¼ cðxÞyðtÞ When this substitution is made, we obtain
2 d2 c h dy þ VðxÞcy ¼ i hc y dt 2m dx2
Division of both sides of this equation by cy gives
2 1 d2 c h 1 dy þ VðxÞ ¼ i h y dt 2m c dx2
Only the left-hand side of this equation is a function of x, so when x changes, only the left-hand side can change. But as the left-hand side is equal to the right-hand side, and the latter does not change, the left-hand side must be equal to a constant. Because the dimensions of the constant are those of an energy (the same as those of V), we shall write it E. It follows that the timedependent equation separates into the following two differential equations:
2 d2 c h þ VðxÞc ¼ Ec 2m dx2
ð1:29aÞ
ih
dy ¼ Ey dt
ð1:29bÞ
The second of these equations has the solution y / eiEt=h
ð1:30Þ
Therefore, the complete wavefunction (C ¼ cy) has the form Cðx, tÞ ¼ cðxÞeiEt=h
ð1:31Þ
The constant of proportionality in eqn 1.30 has been absorbed into the normalization constant for c. The time-independent wavefunction satisfies eqn 1.29a, which may be written in the form Hc ¼ Ec This expression is the time-independent Schro¨dinger equation, on which much of the following development will be based. This analysis stimulates several remarks. First, eqn 1.29a has the form of a standing-wave equation. Therefore, so long as we are interested only in the spatial dependence of the wavefunction, it is legitimate to regard the timeindependent Schro¨dinger equation as a wave equation. Second, when the potential energy of the system does not depend on the time, and the system is in a state of energy E, it is a very simple matter to construct the timedependent wavefunction from the time-independent wavefunction simply by
1.15 SIMULTANEOUS OBSERVABLES
j
25
multiplying the latter by eiEt/h. The time dependence of such a wavefunction is simply a modulation of its phase, because we can write We have used Euler’s relation, e
ix
hÞ i sinðEt= hÞ eiEt=h ¼ cosðEt=
¼ cos x þ i sin x
as well as sin(x) ¼ sin(x) and cos(x) ¼ cos(x).
Re
It follows that the time-dependent factor oscillates periodically from 1 to i to 1 to i and back to 1 with a frequency E/h and period h/E. This behaviour is depicted in Fig. 1.1. Therefore, to imagine the time-variation of a wavefunction of a definite energy, think of it as flickering from positive through imaginary to negative amplitudes with a frequency proportional to the energy. Although the phase of a wavefunction C with definite energy E oscillates in time, the product C C (or jCj2) remains constant: C C ¼ ðc eiEt=h ÞðceiEt=h Þ ¼ c c
x
Im Fig. 1.1 A wavefunction corresponding to an energy E rotates in the complex plane from real to imaginary and back to real at a circular frequency E/ h.
States of this kind are called stationary states. From what we have seen so far, it follows that systems with a specific, precise energy and in which the potential energy does not vary with time are in stationary states. Although their wavefunctions flicker from one phase to another in repetitive manner, the value of C C remains constant in time.
The specification and evolution of states Let us suppose for the moment that the state of a system can be specified as ja,b, . . . i, where each of the eigenvalues a, b, . . . corresponds to the operators representing different observables A, B, . . . of the system. If the system is in the state ja,b, . . . i, then when we measure the property A we shall get exactly a as an outcome, and likewise for the other properties. But can a state be specified arbitrarily fully? That is, can it be simultaneously an eigenstate of all possible observables A, B, . . . without restriction? With this question we are moving into the domain of the uncertainty principle.
1.15 Simultaneous observables As a first step, we establish the conditions under which two observables may be specified simultaneously with arbitrary precision. That is, we establish the conditions for a state jci corresponding to the wavefunction c to be simultaneously an eigenstate of two operators A and B. In fact, we shall prove the following: Property 3. If two observables are to have simultaneously precisely defined values, then their corresponding operators must commute. That is, AB must equal BA, or equivalently, [A,B] ¼ 0. Proof 1.3 Simultaneous eigenstates
Assume that jci is an eigenstate of both operators: Ajci ¼ ajci and Bjci ¼ bjci. That being so, we can write the following chain of relations: ABjci ¼ Abjci ¼ bAjci ¼ bajci ¼ abjci ¼ aBjci ¼ Bajci ¼ BAjci
26
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
Therefore, if jci is an eigenstate of both A and B, and if the same is true for all functions c of a complete set, then it is certainly necessary that [A,B] ¼ 0. However, does the condition [A,B] ¼ 0 actually guarantee that A and B have simultaneous eigenvalues? In other words, if Ajci ¼ ajci and [A,B] ¼ 0, can we be confident that jci is also an eigenstate of B? We confirm this as follows. Because Ajci ¼ ajci, we can write BAjci ¼ Bajci ¼ aBjci Because A and B commute, the first term on the left is equal to ABjci. Therefore, this relation has the form AðBjciÞ ¼ aðBjciÞ However, on comparison of this eigenvalue equation with Ajci ¼ ajci, we can conclude that Bjci / jci, or Bjci ¼ bjci, where b is a coefficient of proportionality. That is, jci is an eigenstate of B, as was to be proved.
It follows from this discussion that we are now in a position to determine which observables may be specified simultaneously. All we need do is to inspect the commutator [A,B]: if it is zero, then A and B may be specified simultaneously. Example 1.7 How to decide whether observables may be specified simultaneously
What restrictions are there on the simultaneous specification of the position and the linear momentum of a particle? Method. To answer this question we have to determine whether the position
coordinates can be specified simultaneously, whether the momentum components can be specified simultaneously, and whether the position and momentum can be specified simultaneously. The answer is found by examining the commutators (Section 1.10; Postulate 2) of the corresponding operators.
x pz
py
y
z px
Fig. 1.2 A summary of the position
and momentum observables that can be specified simultaneously with arbitrary precision (joined by solid lines) and those that cannot (joined by dotted lines).
Answer. All three position operators x, y, and z commute with one another, so there is no constraint on the complete specification of position. The same is true of the three operators for the components of linear momentum. So all three components can be determined simultaneously. However, x and px do not commute, so these two observables cannot be specified simultaneously, and likewise for (y,py) and (z,pz). The consequent pattern of permitted simultaneous specifications is illustrated in Fig. 1.2. Self-test 1.7. Can the kinetic energy and the linear momentum be specified
simultaneously? [Yes]
Pairs of observables that cannot be determined simultaneously are said to be complementary. Thus, position along the x-axis and linear momentum
1.16 THE UNCERTAINTY PRINCIPLE
j
27
parallel to that axis are complementary observables. Classical physics made the mistake of presuming that there was no restriction on the simultaneous determination of observables, that there was no complementarity. Quantum mechanics forces us to choose a selection of all possible observables if we seek to specify a state fully.
1.16 The uncertainty principle Although we cannot specify the eigenvalues of two non-commuting operators simultaneously, it is possible to give up precision in the specification of one property in order to acquire greater precision in the specification of a complementary property. For example, if we know the location of a particle to within a range Dx, then we can specify the linear momentum parallel to x to within a range Dpx subject to the constraint DxDpx 12 h
ð1:32Þ
Thus, as Dx increases (an increased uncertainty in x), the uncertainty in px can decrease, and vice versa. This relation between the uncertainties in the specification of two complementary observables is a special case of the uncertainty principle proposed by Werner Heisenberg in 1927. A very general form of the uncertainty principle was developed by H.P. Robertson in 1929 for two observables A and B: DADB 12 jh½A, B ij
ð1:33Þ
where the root mean square deviation of A is defined as n o1=2 DA ¼ hA2 i hAi2
ð1:34Þ
This is an exact and precise form of the uncertainty principle: the precise form of the ‘uncertainties’ DA and DB are given (they are root mean square deviations) and the right-hand side of eqn 1.33 gives a precise lower bound on the value of the product of uncertainties. Proof 1.4 The uncertainty principle
Suppose that the observables A and B obey the commutation relation [A,B] ¼ iC. (The imaginary i is included for future convenience. For A ¼ x and h.) We B ¼ px it follows from the fundamental commutation relation that C ¼ shall suppose that the system is prepared in a normalized but otherwise arbitrary state jci, which is not necessarily an eigenstate of either operator A or B. The mean values of the observables A and B are expressed by the expectation values hAi ¼ hcjAjci and hBi ¼ hcjBjci The operators for the spread of individual determinations of A and B around their mean values are dA ¼ A hAi and
dB ¼ B hBi
28
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
It is easy to verify that the commutation relation for these deviation operators is ½dA, dB ¼ ½A hAi, B hBi ¼ ½A, B ¼ iC because the expectation values hAi and hBi are simple numbers and commute with operators. Now consider the properties of the following integral, where a is a real but otherwise arbitrary number: Z I ¼ jðadA idBÞcj2 dt The integral I is clearly non-negative as the integrand is positive everywhere. This integral can be developed as follows: Z I ¼ fða dA idBÞcg fða dA idBÞcg dt Z ¼ c ðadA þ idBÞðadA idBÞc dt In the second step we have used the hermitian character of the two operators (as expressed in eqn 1.21b). At this point it is convenient to recognize that the final expression is an expectation value, and to write it in the form I ¼ hðadA þ idBÞðadA idBÞi This expression expands to I ¼ a2 hðdAÞ2 i þ hðdBÞ2 i iahdAdB dBdAi ¼ a2 hðdAÞ2 i þ hðdBÞ2 i þ ahCi In the second step we have recognized the presence of the commutator. The integral is still non-negative, even though that is no longer obvious. At this point we recognize that I has the general form of a quadratic expression in a, and so express it as a square: !2 hCi hCi2 2 aþ þ ðdBÞ2 I ¼ ðdAÞ 2 4 ðdAÞ2 2 ðdAÞ (We have ‘completed the square’ for the first term.) This expression is still nonnegative whatever the value of a, and remains non-negative even if we choose a value for a that corresponds to the minimum value of I. That value of a is the value that ensures that the first term on the right is zero (because that term always supplies a positive contribution to I). Therefore, with that choice of a, we obtain hCi2 I ¼ ðdBÞ2 0 4 ðdAÞ2 The inequality rearranges to ðdAÞ2 hðdBÞ2 14 hCi2 The expectation values on the left can be put into a simpler form by writing them as follows: ðdAÞ2 ¼ hðA hAiÞ2 i ¼ hA2 2AhAi þ hAi2 i ¼ hA2 i 2hAihAi þ hAi2 ¼ hA2 i hAi2
1.17 CONSEQUENCES OF THE UNCERTAINTY PRINCIPLE
j
29
We see that h(dA)2i is the mean square deviation of A from its mean value (and likewise for B). Then the inequality becomes DADB 12 jhCij Then, because [A, B] ¼ iC, we obtain the final form of the uncertainty principle in eqn 1.33.
1.17 Consequences of the uncertainty principle The first point to note is that the uncertainty principle is consistent with Property 3, for if A and B commute, then C is zero and there is no constraint on the uncertainties: there is no inconsistency in having both DA ¼ 0 and DB ¼ 0. On the other hand, when A and B do not commute, the values of DA and DB are related. For instance, while it may be possible to prepare a system in a state in which DA ¼ 0, the uncertainty then implies that DB must be infinite in order to ensure that DADB is not less than 12jh[A,B]ij. In the particular case of the simultaneous specification of x and px, as we have seen, [x, px] ¼ ih, so the lower bound on the simultaneous specification of these two h. complementary observables is 12 Example 1.8 How to calculate the joint uncertainty in two observables
A particle was prepared in a state with wavefunction c ¼ N exp( x2/2G), where N ¼ (1/pG)1/4. Evaluate Dx and Dpx, and confirm that the uncertainty principle is satisfied. Method. We must evaluate the expectation values hxi, hx2i, hpxi, and hp2xi by
integration and then combine their values to obtain Dx and Dpx. There are two short cuts. For hxi, we note that c is symmetrical around x ¼ 0, and so hxi ¼ 0. The value of hpxi can be obtained by noting that px is an imaginary hermitian operator and c is real. Because hermiticity implies that hpxi ¼ hpxi whereas the imaginary character of px implies that hpxi ¼ hpxi, we can conclude that hpxi ¼ 0. For the remaining integrals we use Z 1 Z 1 p1=2 1 p1=2 2 2 eax dx ¼ and x2 eax dx ¼ a 2a a 1 1 Answer. The following integrals are obtained:
hx2 i ¼ N2
Z
1
2
x2 ex
=G
1
hp2x i ¼ N2
Z
1
1
¼ h2 N 2
2
ex
=2G
1 dx ¼ G 2 ! 2 2 2 d ex =2G dx h dx2
Z 1 Z 1 1 1 2 x2 =G h2 2 ex =G dx 2 x e dx ¼ G 1 2G G 1
30
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
It follows that (because hxi ¼ 0 and hpxi ¼ 0) DxDpx ¼ hx2 i1=2 hp2x i1=2 ¼ 12 h Comment. In this example, DxDpx has its minimum permitted value. This is a
special feature of ‘gaussian’ wavefunctions, wavefunctions of the form exp(ax2). A gaussian wavefunction is encountered in the ground state of a harmonic oscillator (see Section 2.16). Self-test 1.8. Calculate the value of DxDpx for a wavefunction that is zero
everywhere except in a region of space of length L, where it has the form (2/L)1/2 sin(px/L). [( h/2(3)1/2)(p2 6)1/2]
The uncertainty principle in the form given in eqn 1.33 can be applied to all pairs of complementary observables. We shall see additional examples in later chapters.
1.18 The uncertainty in energy and time Finally, it is appropriate at this point to make a few remarks about the so-called energy–time uncertainty relation, which is often expressed in the form DEDt h and interpreted as implying a complementarity between energy and time. As we have seen, for this relation to be a true uncertainty relation, it would be necessary for there to be a non-zero commutator for energy and time. However, although the energy operator is well defined (it is the hamiltonian for the system), there is no operator for time in quantum mechanics. Time is a parameter, not an observable. Therefore, strictly speaking, there is no uncertainty relation between energy and time. In Section 6.18 we shall see the true significance of the energy–time ‘uncertainty principle’ is that it is a relation between the uncertainty in the energy of a system that has a finite lifetime t (tau), and is of the form dE h/2t.
1.19 Time-evolution and conservation laws As well as determining which operators are complementary, the commutator of two operators also plays a role in determining the time-evolution of systems and in particular the time-evolution of the expectation values of observables. The precise relation for operators that do not have an intrinsic dependence on the time (in the sense that qO/qt ¼ 0) is dhOi i ¼ h½H, O i dt h
ð1:35Þ
We see that if the operator for the observable commutes with the hamiltonian, then the expectation value of the operator does not change with time. An observable that commutes with the hamiltonian for the system, and which therefore has an expectation value that does not change with time, is called a constant of the motion, and its expectation value is said to be conserved.
1.19 TIME-EVOLUTION AND CONSERVATION LAWS
j
31
Proof 1.5 Time evolution
Differentiation of hOi with respect to time gives Z Z dhOi d qC qC OC dt þ C O ¼ hCjOjCi ¼ dt qt dt dt qt because only the state C (not the operator O) depends on the time. The Schro¨dinger equation lets us write Z Z Z qC 1 1 dt ¼ C O HC dt ¼ C OHC dt C O qt i h i h Z
Z Z qC 1 1 ðHCÞ OC dt ¼ C HOC dt OC dt ¼ i h i h qt
In the second line we have used the hermiticity of the hamiltonian (in the form of eqn 1.21b). It then follows, by combining these two expressions, that dhOi 1 i ¼ ðhHOi hOHiÞ ¼ h½H, O i dt ih h as was to be proved.
As an important example, consider the rate of change of the expectation value of the linear momentum of a particle in a one-dimensional system. The commutator of H and px is " # h2 d2 h d h d ¼ V, þ V, ½H, px ¼ i dx i dx 2m dx2 because the derivatives commute. The remaining commutator can be evaluated by remembering that there is an unwritten function on the right on which the operators operate, and writing h dc dðVcÞ h dc dc dV ½H, px c ¼ V ¼ V V c i dx dx i dx dx dx ¼
dV h c i dx
This relation is true for all functions c; therefore the commutator itself is ½H, px ¼
dV h i dx
ð1:36Þ
It follows that the linear momentum is a constant of the motion if the potential energy does not vary with position, that is when dV/dx ¼ 0. Specifically, we can conclude that the rate of change of the expectation value of linear momentum is d i dV hpx i ¼ h½H, px i ¼ ð1:37Þ dt h dx
32
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
Then, because the negative slope of the potential energy is by definition the force that is acting (F ¼ dV/dx), the rate of change of the expectation value of linear momentum is given by d hpx i ¼ hFi dt
ð1:38Þ
That is, the rate of change of the expectation value of the linear momentum is equal to the expectation value of the force. It is also quite easy to prove in the same way that d hpx i hxi ¼ dt m
ð1:39Þ
which shows that the rate of change of the mean position can be identified with the mean velocity along the x-axis. These two relations jointly constitute Ehrenfest’s theorem. Ehrenfest’s theorem clarifies the relation between classical and quantum mechanics: classical mechanics deals with average values (expectation values); quantum mechanics deals with the underlying details.
Matrices in quantum mechanics As we have seen, the fundamental commutation relation of quantum mechanics, [x,px] ¼ ih, implies that x and px are to be treated as operators. However, there is an alternative interpretation: that x and px should be represented by matrices, for matrix multiplication is also non-commutative. We shall introduce this approach here as it introduces a language that is widely used throughout quantum mechanics even though matrices are not being used explicitly.
1.20 Matrix elements A matrix, M, is an array of numbers (which may be complex), called matrix elements. Each element is specified by quoting the row (r) and column (c) that it occupies, and denoting the matrix element as Mrc. The rules of matrix algebra are set out in Further information 23. For our present purposes it is sufficient to emphasize the rule of matrix multiplication: the product of two matrices M and N is another matrix P ¼ MN with elements given by the rule X Mrs Nsc ð1:40Þ Prc ¼ s
The order of matrix multiplication is important, and it is essential to note that MN is not necessarily equal to NM. Hence, MN NM is not in general zero. Heisenberg formulated his version of quantum mechanics, which is called matrix mechanics, by representing position and linear momentum by the matrices x and px, and requiring that xpx pxx ¼ ih1 where 1 is the unit matrix, a square matrix with all diagonal elements (those for which r ¼ c) equal to 1 and all others 0.
1.20 MATRIX ELEMENTS
j
33
Throughout this chapter we have encountered quantities of the form hmjOjni. These quantities are commonly abbreviated as Omn, which immediately suggests that they are elements of a matrix. For this reason, the Dirac bracket hmjOjni is often called a matrix element of the operator O. A diagonal matrix element Onn is then a bracket of the form hnjOjni with the bra and the ket referring to the same state. We shall often encounter sums over products of Dirac brackets that have the form X hrjAjsihsjBjci s
If the brackets that appear in this expression are interpreted as matrix elements, then we see that it has the form of a matrix multiplication, and we may write X X hrjAjsihsjBjci ¼ Ars Bsc ¼ ðABÞrc ¼ hrjABjci ð1:41Þ s
s
That is, the sum is equal to the single matrix element (bracket) of the product of operators AB. Comparison of the first and last terms in this line of equations also allows us to write the symbolic relation X jsihsj ¼ 1 ð1:42Þ s
This completeness relation is exceptionally useful for developing quantum mechanical equations. It is often used in reverse: the matrix element hrjABjci can always be split into a sum of two factors by regarding it as hrjA1Bjci and then replacing the 1 by a sum over a complete set of states of the form in eqn 1.42. Example 1.9 How to make use of the completeness relation
Use the completeness relation to prove that the eigenvalues of the square of an hermitian operator are non-negative. Method. We have to prove, for O2joi ¼ ojoi, that o 0 if O is hermitian.
If both sides of the eigenvalue equation are multiplied by hoj, converting it to hojO2joi ¼ o, we see that the proof requires us to show that the expectation value on the left is non-negative. As it has the form hojOOjoi, it suggests that the completeness relation might provide a way forward. The hermiticity of O implies that it will be appropriate to use the property hmjOjni ¼ hnjOjmi at some stage in the argument. Answer. The diagonal matrix element hojO2joi can be developed as follows:
hojO2 joi ¼ hojOOjoi ¼
X hojOjsihsjOjoi s
¼
X X hojOjsihojOjsi ¼ jhojOjsij2 0 s
s
The final inequality follows from the fact that all the terms in the sum are non-negative. Self-test 1.9. Show that if (Of ) ¼ Of , then hOi ¼ 0 for any real function f.
34
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
The origin of the completeness relation, which is also known as the closure relation, can be demonstrated by the following argument. Suppose we have a complete set of orthonormal states jsii. Then, by definition of complete, we can expand an arbitrary state jci as a linear combination: X jci ¼ ci jsi i i
Multiplication from the left by the bra hsjj and use of the orthonormality of the complete basis set gives cj ¼ hsjjci. Thus X X jci ¼ hsi jcijsi i ¼ jsi ihsi jci i
i
which immediately implies the completeness relation.
1.21 The diagonalization of the hamiltonian The time-independent form of the Schro¨dinger equation, Hc ¼ Ec, can be given a matrix interpretation. First, we express jci as a linear combination of a complete set of states jni: X X Hjci ¼ H cn jni ¼ cn Hjni n
Ejci ¼ E
X
n
cn jni
n
These two lines are equal to one another. Next, multiply the right-hand sides of the above two equations from the left by an arbitrary bra hmj and use the orthonormality of the states to obtain X X cn hmjHjni ¼ E cn hmjni ¼ Ecm n
n
In matrix notation this equation is X Hmn cn ¼ Ecm
ð1:43Þ
n
Now suppose that we can find the set of states such that Hmn ¼ 0 unless m ¼ n; that is, when using this set, the hamiltonian has a diagonal matrix. Then this expression becomes Hmm cm ¼ Ecm
ð1:44Þ
and the energy E is seen to be the diagonal element of the hamiltonian matrix. In other words, solving the Schro¨dinger equation is equivalent to diagonalizing the hamiltonian matrix (see Further information 23). This is yet another link between the Schro¨dinger and Heisenberg formulations of quantum mechanics. Indeed, it was reported that when Heisenberg was looking for ways of diagonalizing his matrices, the mathematician David Hilbert suggested to him that he should look for the corresponding differential equation instead. Had he done so, Schro¨dinger’s wave mechanics would have been Heisenberg’s too.
1.21 THE DIAGONALIZATION OF THE HAMILTONIAN
j
35
Example 1.10 How to diagonalize a simple hamiltonian
In a system that consists of only two orthonormal states j1i and j2i (such as electron spin in a magnetic field, when the electron spin can be in one of two orientations), the hamiltonian has the following matrix elements: H11 ¼ h1jHj1i ¼ a, H22 ¼ h2jHj2i ¼ b, H12 ¼ d, H21 ¼ d . For notational simplicity, we shall suppose that d is real, so d ¼ d. Find the energy levels and the eigenstates of the system. Method. The energy levels are the eigenvalues of the hamiltonian matrix.
We use the procedure explained in Further information 23 to find the eigenvalues and eigenstates. We describe the procedure here briefly, specifically for the two-state system. One eigenstate is jji ¼ c1j1i þ c2j2i and the other is jki ¼ d1j1i þ d2j2i. Beginning twice with Hjji ¼ Ejji and multiplying one on the left by h1j and the second on the left by h2j, we obtain two equations which in matrix form are c1 H11 E H12 ¼0 H21 H22 E c2 There is a (non-trivial, c1 and c2 non-zero) solution to this matrix equation only if the determinant of the matrix on the left-hand side vanishes. A similar argument develops if we begin with Hjki ¼ Ejki. The two energy eigenvalues are determined from the secular determinant jH E1j ¼ 0 and the two energy eigenvalues, denoted E, are the diagonal elements of the matrix E. To find the eigenstates, we form the matrix T composed of the two column vectors of the eigenstates: c1 d1 T¼ c2 d2 The matrix T satisfies the equation HT ¼ TE. The best procedure is to choose the coefficients c1, c2, d1, and d2 so that the eigenstates are given by jji ¼ j1i cos z þ j2i sin z and jki ¼ j1i sin z þ j2i cos z, where z is a parameter, for this parametrization ensures that the two eigenstates are orthonormal for all values of z. After solving the secular determinant equation for the eigenvalues, we form T1HT, equate it to the matrix E, and then solve for z. Answer. Because the states j1i and j2i are orthonormal, the secular determinant is
a E d
detjH E1j ¼
¼ ða EÞðb EÞ d2 ¼ 0 d b E
This quadratic equation for E has the roots E ¼ 12 ða þ bÞ 12 fða bÞ2 þ 4d2 g1=2 ¼ 12 ða þ bÞ D where D ¼ 12 {(a b)2 þ 4d2}1/2. These are the eigenvalues, and hence they are the energy levels. We next form the transformation matrix and its reciprocal: cos z sin z cos z sin z T¼ T 1 ¼ sin z cos z sin z cos z
36
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
Then construct the following matrix equation:
Eþ 0 cosz sinz cosz sinz ad ¼T 1 HT ¼ 0 E sinz cosz sinz cosz db
a cos2 zþb sin2 zþ2d cosz sinz dðcos2 zsin2 zÞþðbaÞcosz sinz ¼ dðcos2 zsin2 zÞþðbaÞcosz sinz b cos2 zþa sin2 z2d cosz sinz
!
Consequently, by equating matching off-diagonal elements, we obtain dðcos2 z sin2 zÞ þ ðb aÞ cos z sin z ¼ 0 which solves to z ¼ 12 arctan
2d ba
Comment. The two-level system occurs widely in quantum mechanics, and we
shall return to it in Chapter 6. The parametrization of the states in terms of the angle z is a very useful device, and we shall encounter it again.
The plausibility of the Schro¨dinger equation The Schro¨dinger equation is properly regarded as a postulate of quantum mechanics, and hence we should not ask for a deeper justification. However, it is often more satisfying to set postulates in the framework of the familiar. In this section we shall see that the Schro¨dinger equation is a plausible description of the behaviour of matter by going back to the formulation of classical mechanics devised by W.R. Hamilton in the nineteenth century. We shall concentrate on the qualitative aspects of the approach: the calculations supporting these remarks will be found in Further information 1.
1.22 The propagation of light
P1
P2
Fig. 1.3 When light reflects from a surface, the angle of reflection is equal to the angle of incidence.
In geometrical optics, light travels in straight lines in a uniform medium, and we know that the physical nature of light is a wave motion. In classical mechanics particles travel in straight lines unless a force is present. Moreover, we know from the experiments performed at the end of the nineteenth century and the start of the twentieth century that particles have a wave character. There are clearly deep analogies here. We shall therefore first establish how, in optics, wave motion can result in straight-line motion, and then argue by analogy about the wave nature of particles. The basic rule governing light propagation in geometrical optics is Fermat’s principle of least time. A simple form of the principle is that the path taken by a ray of light through a medium is such that its time of passage is a minimum. As an illustration, consider the relation between the angles of incidence and reflection for light falling on a mirror (Fig. 1.3). The briefest path between source, mirror, and observer is clearly the one corresponding to equal angles of incidence and reflection. In the case of refraction, it is necessary to take into
1.22 THE PROPAGATION OF LIGHT P1
i r
P2 Fig. 1.4 When light is refracted at the
interface of two transparent media, the angle of refraction, yr, and the angle of incidence, yi, are related by Snell’s law.
A⬘
j
37
account the different speeds of propagation in the two media. In Fig. 1.4, the geometrically straight path is not necessarily the briefest, because the light travels relatively slowly through the denser medium. The briefest path is in fact easily shown to be the one in which the angles of incidence yi and refraction yr are related by Snell’s law, that sin yr/sin yi ¼ n1/n2. (The refractive indexes n1 and n2 enter because the speed of light in a medium of refractive index n is c/n, where c is the speed of light in a vacuum.) How can the wave nature of light account for this behaviour? Consider the case illustrated in Fig. 1.5, where we are interested in the propagation of light between two fixed points P1 and P2. A wave of electromagnetic radiation travelling along some general path A arrives at P2 with a particular phase that depends on its path length. A wave travelling along a neighbouring path A 0 travels a different distance and arrives with a different phase. Path A has very many neighbouring paths, and there is destructive interference between the waves. Hence, an observer concludes that the light does not travel along a path like A. The same argument applies to every path between the two points, with one exception: the straight line path B. The neighbours of B do not interfere destructively with B itself, and it survives. The mathematical reason for this exceptional behaviour can be seen as follows. The amplitude of a wave at some point x can be written ae2pix/l, where l is the wavelength. It follows that the amplitude at P1 is ae2pix1/l and that at P2 it is ae2pix2/l. The two amplitudes are therefore related as follows: CðP2 Þ ¼ ae2pix2 =l ¼ e2piðx2 x1 Þ=l e2pix1 =l ¼ e2piðx2 x1 Þ=l CðP1 Þ This relation between the two amplitudes can be written more simply as CðP2 Þ ¼ eif CðP1 Þ with f ¼ 2pðx2 x1 Þ=l
(a) A
P1
P2
B⬘
B
(b)
P1
P2 Fig. 1.5 (a) A curved path through a
uniform medium has neighbours with significantly different phases at the destination point, and there is destructive interference between them. (b) A straight path between two points has neighbours with almost the same phase, and these paths do not interfere destructively.
ð1:45Þ
The function f is the phase length of the straight-line path. The relative phases at P2 and P1 for waves that travel by curved paths are related by an expression of the same kind, but with the phase length determined by the length, L, of the path: 2pL ð1:46Þ f¼ l Now we consider how the path length varies with the distortion of the path from a straight line. If we distort the path from B to A in Fig. 1.5, f changes as depicted in Fig. 1.6. Obviously, f goes through a minimum at B. Now we arrive at the crux of the argument. Consider the phase length of the paths in the vicinity of A. The phase length of A 0 is related to the phase length at A by the following Taylor expansion: ! 2 df 0 1 d f ds þ 2 ds2 þ ð1:47Þ fðA Þ ¼ fðAÞ þ ds A ds2 A
where ds is a measure of the distortion of the path. This expression should be compared with the similar expression for ! the path lengths of B and its neighbours: 2 df d f ds þ 12 ds2 þ fðB0 Þ ¼ fðBÞ þ ds B ds2 B ! 2 d f ds2 þ ð1:48Þ ¼ fðBÞ þ 12 ds2 B
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
s
A⬘ A
B B⬘
Phase length,
38
δs δs
B B⬘
A A⬘
Displacement, s
Fig. 1.6 The variation of phase length
Phase length,
with displacement from a straight line path. The phase length at A 0 differs from that at A by a first-order term; the phase lengths at B and B 0 differ only to second order in the displacement.
Decreasing wavelength
Displacement, s Fig. 1.7 The variation of phase length with wavelength. Interference between neighbours is most acute for short wavelengths. The geometrical limit corresponds to zero wavelength, where even infinitesimal neighbours interfere destructively and completely.
The term in ds is zero because the first derivative is zero at the minimum of the curve. In other words, to first order in the displacement, straight line paths have neighbours with the same phase length. On the other hand, curved paths have neighbours with different phase lengths. This difference is the reason why straight line propagation survives whereas curved paths do not: the latter have annihilating neighbours. Two further points now need to be made. When the medium is not uniform, the wavelength of a wave varies with position. Because l ¼ v/n, and v, the speed of propagation, is equal to c/n, where the refractive index n varies with position, a more general form of the phase length is Z Z P2 dx 2pn P2 nðxÞ dx ð1:49Þ ¼ f ¼ 2p c P1 P1 lðxÞ The same argument applies, but because of the dependence of the refractive index on position, a curved or kinked path may turn out to correspond to the minimum phase length, and therefore have, to first order at least, no destructive neighbours. Hence, the path adopted by the light will be curved or kinked. The focusing caused by a lens is a manifestation of this effect. The second point concerns the stringency of the conclusion that the minimum-phase-length paths have non-destructive neighbours. Because the wavelength of the radiation occurs in the denominator of the expression defining the phase length, waves of short wavelength will have larger phase lengths for a given path than radiation of long wavelength. The variation of phase length with wavelength is indicated in Fig. 1.7. It should be clear that neighbours annihilate themselves much more strongly when the light has a short wavelength than when it is long. Therefore, the rule that light (or any other form of wave motion) propagates itself in straight lines becomes more stringent as its wavelength shortens. Sound waves travel only in approximately straight lines; light waves travel in almost exactly straight lines. Geometrical optics is the limit of infinitely short wavelengths, where the annihilation by neighbours is so effective that the light appears to travel in perfectly straight lines.
1.23 The propagation of particles The path taken by a particle in classical mechanics is determined by Newton’s laws. However, it turns out that these laws are equivalent to Hamilton’s principle, which states that particles adopt paths between two given points such that the action S associated with the path is a minimum. There is clearly a striking analogy between Fermat’s principle of least time and Hamilton’s principle of least action. The formal definition of action is given in Further information 1, where it is seen to be an integral taken along the path of the particle, just like the phase length in optics. When we turn to the question of why particles adopt the path of least action, we can hardly avoid the conclusion that the reason must be the same as why light adopts the path of least phase length. But to apply that argument to particles, we have to suppose that particles have an associated wave character. You can see that this attempt to ‘explain’ classical mechanics
1.24 THE TRANSITION TO QUANTUM MECHANICS
j
39
leads almost unavoidably to the heart of quantum mechanics and the duality of matter. We have the experimental evidence to encourage us to pursue the analogy; Hamilton did not.
1.24 The transition to quantum mechanics The hypothesis we now make is that a particle is described by some kind of amplitude C, and that amplitudes at different points are related by an expression of the form C(P2) ¼ eifC(P1). By analogy with optics, we say that the wave is propagated along the path that makes f a minimum. But we also know that in the classical limit, the particle propagates along a path that corresponds to least action. As f is dimensionless (because it appears as an exponent), the constant of proportionality between f and S must have the dimensions of 1/action. Furthermore, we have seen that geometrical optics, the classical form of optics, corresponds to the limit of short wavelengths and very large phase lengths. In classical mechanics, particles travel along ‘geometrical’ trajectories, corresponding to large f. Hence, the constant with the dimensions of action must be very small. The natural quantity to introduce is Planck’s constant, or some small multiple of it. It turns out that agreement with experiment (that is, the correct form of the Schro¨dinger equation) is obtained if we use h; we therefore conclude that we should write f ¼ S/h. You should notice the relation between this approach and Heisenberg’s. In his, a 0 was replaced by h (in the commutator [x,px]), and classical mechanics ‘evolved’ into quantum mechanics. In the approach we are presenting here, a 0 has also been replaced by h, for had we wanted precise geometrical trajectories, then we would have divided S by 0. We have arrived at the stage where the amplitude associated with a particle is described by a relation of the form CðP2 Þ ¼ eiS=h CðP1 Þ
ð1:50Þ
where S is the action associated with the path from P1 (at x1, t1) to P2 (at x2, t2). This expression lets us develop an equation of motion, because we can differentiate C with respect to the time t2: qCðP2 Þ i qS iS=h i qS ¼ e CðP1 Þ ¼ CðP2 Þ qt2 h qt2 h qt2 One of the results derived in Further information 1 is that the rate of change of the action is equal to E, where E is the total energy, T þ V: qS ¼ E ð1:51Þ qt Therefore, the equation of motion at all points of a trajectory is qC i ¼ EC qt h The final step involves replacing E by its corresponding operator H, which then results in the time-dependent Schro¨dinger equation, eqn 1.27.
40
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS
There are a few points that are worth noting about this justification. First, we have argued by analogy with classical optics, and have sought to formulate equations that are consistent with classical mechanics. It should therefore not be surprising that the approach might not generate some purely quantum mechanical properties. Indeed, we shall see later that the property of electron spin has been missed, for despite its evocative name, spin has no classical counterpart. A related point is that the derivation has been entirely non-relativistic: at no point have we tried to ensure that space and time are treated on an equal footing. The alignment of relativity and quantum mechanics was achieved by P.A.M. Dirac, who found a way of treating space and time symmetrically, and in the process accounted for the existence of electron spin. Finally, it should be noted that the time-dependent Schro¨dinger equation is not a wave equation. A wave equation has a second derivative with respect to time, whereas the Schro¨dinger equation has a first derivative. We have to conclude that the time-dependent Schro¨dinger equation is therefore a kind of diffusion equation, an equation of the form qf ¼ Dr2 f qt
ð1:52Þ
where f is a probability density and D is a diffusion coefficient. There is perhaps an intuitive satisfaction in the notion that the solutions of the basic equation of quantum mechanics evolve by some kind of diffusion.
PROBLEMS 1.1 Which of the following operations are linear and which are non-linear: (a) integration, (b) extraction of a square root, (c) translation (replacement of x by x þ a, where a is a constant), (d) inversion (replacement of x by x)? 1.2 Find the operator for position x if the operator for momentum p is taken to be ( h/2m)1/2(A þ B), with [A,B] ¼ 1 and all other commutators zero. 1.3 Which of the following functions are eigenfunctions of 2 (a) d/dx, (b) d2/dx2: (i) eax, (ii) eax , (iii) x, (iv) x2, (v) ax þ b, (vi) sin x? 1.4 Construct quantum mechanical operators in the position representation for the following observables: (a) kinetic energy in one and in three dimensions, (b) the inverse separation, 1/x, (c) electric dipole moment, (d) z-component of angular momentum, (e) the mean square deviations of the position and momentum of a particle from the mean values. 1.5 Repeat Problem 1.4, but find operators in the momentum representation. Hint. The observable 1/x should be regarded as x1; hence the operator required is the inverse of the operator for x.
1.6 In relativistic mechanics, energy and momentum are related by the expression E2 ¼ p2c2 þ m2c4. Show that when p2c2 m2c4 this expression reduces to E ¼ p2/2m þ mc2. Construct the relativistic analogue of the Schro¨dinger equation from the relativistic expression. What can be said about the conservation of probability? Hint: For the latter part, see Problem 1.36. 1.7 Confirm that the operators (a) T ¼ ( h2/2m)(d2/dx2) and (b) lzR¼ ( h/i)(d/df) are hermitian. Hint. Consider the R 2p L integrals 0 ca Tcb dx and 0 ca lzcb df and integrate by parts. 1.8 Demonstrate that the linear combinations A þ iB and A iB are not hermitian if A and B are hermitian operators. 1.9 Evaluate the expectation values of the operators px and p2x for a particle with wavefunction (2/L)1/2 sin (px/L) in the range 0 to L. 1.10 Are the linear combinations 2x y z, 2y x z, 2z x y linearly independent or not? 1.11 Evaluate the commutators (a) [x,y], (b) [px,py], (c) [x,px], (d) [x2,px], (e) [xn,px].
PROBLEMS
j
41
1.12 Evaluate the commutators (a) [(1/x),px], (b) [(1/x), px2], (c) [xpy ypx, ypz zpy], (d) [x2(q2/qy2), y(q/qx)].
a particle on a ring with uniform potential energy V(f) ¼ V.
1.13 Show that (a) [A,B] ¼ [B,A], (b) [Am,An] ¼ 0 for all m, n, (c) [A2,B] ¼ A[A,B] þ [A,B]A, (d) [A,[B,C] ] þ [B,[C,A] ] þ [C,[A,B] ] ¼ 0.
1.23 The only non-zero matrix elements of x and px for a harmonic oscillator are
1.14 Evaluate the commutator [ly,[ly,lz] ] given that [lx,ly] ¼ ihlz, [ly,lz] ¼ ihlx, and [lz,lx] ¼ ihly. 1.15 The operator eA has a meaning if it is expanded as a power series: eA ¼ Sn(1/n!)An. Show that if jai is an eigenstate of A with eigenvalue a, then it is also an eigenstate of eA. Find the latter’s eigenvalue. 1.16 (a) Show that eAeB ¼ eAþB only if [A,B] ¼ 0. (b) If [A,B] 6¼ 0 but [A,[A,B] ] ¼ [B,[A,B] ] ¼ 0, show that eAeB ¼ eAþBef, where f is a simple function of [A,B]. Hint. This is another example of the differences between operators (q-numbers) and ordinary numbers (c-numbers). The simplest approach is to expand the exponentials and to collect and compare terms on both sides of the equality. Note that eAeB will give terms like 2AB while eAþB will give AB þ BA. Be careful with order. 1.17 Evaluate the commutators (a) [H,px] and (b) [H,x], where H ¼ px2/2m þ V(x). Choose (i) V(x) ¼ V, a constant, (ii) V(x) ¼ 12kx2, (iii) V(x) ! V(r) ¼ e2/4pe0r. 1.18 Evaluate (by considering eqn 1.33) the limitation on the simultaneous specification of the following observables: (a) the position and momentum of a particle, (b) the three components of linear momentum of a particle, (c) the kinetic energy and potential energy of a particle, (d) the electric dipole moment and the total energy of a one-dimensional system, (e) the kinetic energy and the position of a particle in one dimension. 1.19 An electron is confined to a linear box of length 0.10 nm. What are the minimum uncertainties in (a) its velocity and (b) its kinetic energy? 1.20 Use the uncertainty principle to estimate the order of magnitude of the diameter of an atom. Compare the result with the radius of the first Bohr orbit of hydrogen, a0 ¼ 4pe0h2/mee2. Hint. Suppose the electron is confined to a region of extent Dx; this confinement implies a non-zero kinetic energy. There is also a potential energy of order of magnitude e2/4pe0Dx. Find Dx such that the total energy is a minimum, and evaluate the expression.
h 1=2 ðv þ 1Þ1=2 2mo h 1=2 1=2 hv 1jxjvi ¼ v 2mo 1=2 hmo hv þ 1jpx jvi ¼ i ðv þ 1Þ1=2 2 hmo 1=2 1=2 hv 1jpx jvi ¼ i v 2 hv þ 1jxjvi ¼
(and their hermitian conjugates); see Section 2.17. Write out the matrices of x and px explicitly (label the rows and columns v ¼ 0, 1, 2, . . . ) up to about v ¼ 4, and confirm by matrix multiplication that they satisfy the commutation rule. Construct the hamiltonian matrix by forming p2x/2m þ 12kx2 by matrix multiplication and addition, and infer the eigenvalues. 1.24 Use the completeness relation, eqn 1.42, and the information in Problem 1.23 to deduce the value of the matrix element hvjxpx2xjvi. 1.25 Write the time-independent Schro¨dinger equations for (a) the hydrogen atom, (b) the helium atom, (c) the hydrogen molecule, (d) a free particle, (e) a particle subjected to a constant, uniform force. 1.26 The time-dependent Schro¨dinger equation is separable when V is independent of time. (a) Show that it is also separable when V is a function only of time and uniform in space. (b) Solve the pair of equations. Let V(t) ¼ V cos ot; find an expression for C(x, t) in terms of C(x, 0). (c) Is C(x, t) stationary in the sense specified in Section 1.12? 1.27 The ground-state wavefunction of a hydrogen atom has the form c(r) ¼ Nebr, b being a collection of fundamental constants with the magnitude 1/(53 pm). Normalize this spherically symmetrical function. Hint. The volume element is dt ¼ sin y dy df r2 dr, with 0 y p, 0 f 2p, and 0 r < 1. ‘Normalize’ always means ‘normalize to 1’ in this text.
1.21 Use eqn 1.35 to find expressions for the rate of change of the expectation values of position and momentum of a harmonic oscillator; solve the pair of differential equations, and show that the expectation values change in time in the same way as for a classical oscillator.
1.28 A particle in an infinite one-dimensional system was 2 2 described by the wavefunction c(x) ¼ Nex =2G . Normalize this function. Calculate the probability of finding the particle in the range G x G. Hint. The integral encountered in the second part is the error function. It is defined and tabulated in M. Abramowitz and I.A. Stegun, Handbook of mathematical functions, Dover (1965).
1.22 Confirm that the z-component of angular momentum, lz ¼ (h/i) d/df, is a constant of the motion for
1.29 An excited state of the system in the previous problem is described by the wavefunction
42
j
1 THE FOUNDATIONS OF QUANTUM MECHANICS 2
2
cðxÞ ¼ Nxex =2G . Where is the most probable location of the particle? 1.30 On the basis of the information in Problem 1.27, calculate the probability density of finding the electron (a) at the nucleus, (b) at a point in space 53 pm from the nucleus. Calculate the probabilities of finding the electron inside a region of volume 1.0 pm3 located at these points assuming that the probability density is constant inside the small volume region. 1.31 (a) Calculate the probability of the electron being found anywhere within a sphere of radius 53 pm for the atom defined in Problem 1.27. (b) If the radius of the atom is defined as the radius of the sphere inside which there is a 90 per cent probability of finding the electron, what is the atom’s radius? 1.32 A particle is confined to the region 0 x 1 and its state is described by the unnormalized wavefunction c(x) ¼ e2x. What is the probability of finding the particle at a distance x 1? 1.33 A particle is moving in a circle in the xy plane. The only coordinate of importance is the angle f which can vary from 0 to 2p as the particle goes around the circle. We are interested in measurements of the angular momentum L of the particle. The angular momentum operator for such a system is given by (h/i) d/df. (a) Suppose that the state of the particle is described by the wavefunction c(f) ¼ Neif where N is the normalization constant. What values will we find when we measure the angular momentum of the particle? If more than one
value is possible, what is the probability of obtaining each result? What is the expectation value of the angular momentum? (b) Now suppose that the state of the particle is described by the normalized wavefunction c(f) ¼ N{(3/4)1/2eif (i/2)e2if}. When we measure the angular momentum of the particle, what value(s) will we find? If more than one value is possible, what is the probability of obtaining each result? What is the expectation value of the angular momentum? 1.34 Explore the concept of phase length as follows. First, consider two points P1 and P2 separated by a distance l, and let the paths taken by waves of wavelength l be a straight line from P1 to a point a distance d above the midpoint of the line P1P2, and then on to P2. Find an expression for the phase length and sketch it as a function of d for various values of l. Confirm explicitly that f 0 ¼ 0 at d ¼ 0. 1.35 Confirm that the path of minimum phase length for light passing from one medium to another corresponds to light being refracted at their interface in accord with Snell’s law (Section 1.21). 1.36 Show that if the Schro¨dinger equation had the form of a true wave equation, then the integrated probability would be time-dependent. Hint. A wave equation has kq2/qt2 in place of q/qt, where k is a constant with the appropriate dimensions (what are they?). Solve the time component ofR the separable equation and investigate the behaviour of C C dt.
2 The characteristics of acceptable wavefunctions Some general remarks on the Schro¨dinger equation 2.1 The curvature of the wavefunction 2.2 Qualitative solutions 2.3 The emergence of quantization 2.4 Penetration into non-classical regions
Linear motion and the harmonic oscillator
In this chapter we consider the quantum mechanics of translation and vibration. Both types of motion can be solved exactly in certain cases, and both are important not only in their own right but also because they form a basis for the description of the more complicated types of motion encountered in quantum chemistry. Translational motion also has the advantage of introducing in a simple way many of the striking features of quantum mechanics. However, there are certain features of wavefunctions that are common to all the problems we shall encounter, and we start by considering them. As we shall see, it is the combination of these features with the solution of the Schro¨dinger equation that results in one of the most characteristic features of quantum mechanics, the quantization of energy.
Translational motion 2.5 Energy and momentum 2.6 The significance of the coefficients 2.7 The flux density 2.8 Wavepackets Penetration into and through barriers 2.9 An infinitely thick potential wall 2.10 A barrier of finite width 2.11 The Eckart potential barrier Particle in a box 2.12 The solutions 2.13 Features of the solutions 2.14 The two-dimensional square well 2.15 Degeneracy The harmonic oscillator 2.16 The solutions 2.17 Properties of the solutions 2.18 The classical limit Translation revisited: The scattering matrix
The characteristics of acceptable wavefunctions We have seen that the Born interpretation of the wavefunction c, like that of its time-dependent version C, is that c c is a probability density. It must therefore be square-integrable (Section 1.12), and specifically the wavefunction must satisfy the normalization condition Z c c dt ¼ 1 ð2:1Þ The implication of this condition is that the wavefunction cannot become infinite over a finite region of space, as in Fig. 2.1. If it did become infinite, the integral would be infinite, and the Born interpretation would be untenable. This restriction does not rule out the possibility that the wavefunction could be infinite over an infinitesimal region of space because then its integral may remain finite (the integral is the area under the curve of c c, and infinitely high infinitely narrow may result in a finite area). Such a wavefunction corresponds to the localization of a particle at a single, precise point, like the centre of mass of a speck of dust on a table at absolute zero. By the uncertainty principle, we know that a particle described by a wavefunction of this kind would have an infinitely uncertain linear momentum. Another implication of the Born interpretation is that for c c to be a valid probability density, it must be single valued; that is, have one value at each point. The Born interpretation would be untenable if c c could take more than one value at each point of space. In simple applications, the single-valued
∞
Wavefunction,
(b)
x
(c)
x
x
Fig. 2.2 Three unacceptable wavefunctions. (a) A wavefunction that is not single-valued everywhere. (b) A discontinuous wavefunction. (c) A wavefunction with a discontinuous slope.
x ∞ Wavefunction,
(b)
x Fig. 2.1 (a) A wavefunction must not be infinite over a finite range because it is then not squareintegrable. (b) However, it may be infinite over an infinitesimal range for such a function is square-integrable (it corresponds to a Dirac d-function).
Wavefunction,
(a)
Wavefunction,
∞ (a)
Wavefunction,
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
Wavefunction,
j
44
Positive curvature
0
x Zero curvature Negative curvature
Fig. 2.3 The variation of the curvature of a wavefunction with its amplitude, for a constant energy E < V.
character of c c implies that c itself must be single valued, and we shall normally impose that condition on the wavefunction. (The exceptions arise when electron spin is taken into account.) There are two other conditions on the form of the wavefunction that stem from the requirement that c is a solution of a second-order differential equation, and therefore that its second derivative should exist. In the first place, in order to define a second derivative of a function, it is necessary that the function itself should be continuous (Fig. 2.2). A weaker requirement is that the first derivative should also be continuous. This condition is weaker because there are systems—those with certain ill-behaved potential energies— where the restriction is too severe. For example, when we deal with a particle in a box, we encounter a potential energy that is excessively ill-behaved because it jumps from zero to infinity in an infinitesimal distance (when the particle touches the wall of the box). In such a case there is no need for the particle to have a continuous first derivative. In summary, in general a wavefunction must satisfy the following conditions: 1. 2. 3. 4.
Single valued (strictly, c c should be single valued). Not infinite over a finite range. Continuous everywhere. Possess a continuous first derivative, except at ill-behaved regions of the potential.
Some general remarks on the Schro¨dinger equation The time-independent Schro¨dinger equation is an equation for the second derivative of the wavefunction, which we can interpret informally as its curvature. With this idea established, it is possible to guess the form of its solutions even when the form of the potential energy is complicated. and one with negative A function with positive curvature looks like curvature looks like . The one-dimensional Schro¨dinger equation expresses the curvature of the wavefunction as d2 c 2m ¼ 2 ðV EÞc dx2 h
ð2:2Þ
2.2 QUALITATIVE SOLUTIONS
This use of the term curvature is colloquial. In fact, in mathematics, curvature is a precisely defined concept in the theory of surfaces: in one dimension the curvature of a function f is Curvature of f ¼
f1 þ ðdf =dxÞ2 g3=2
f=x 2 d2f/dx2 Curvature
0 –2
0 x
2
For example, the curvature of the parabola f ¼ x2 is 2/(1 þ 4x2)3/2, and decreases as jxj increases, whereas d2f/dx2 ¼ 2, a constant at all values of x (see the illustration). For simplicity of expression, we shall adopt the colloquial meaning, and identify curvature with the second derivative d2f/dx2. E
45
Therefore, if we know the values of V E and c at a particular point, then we can state the curvature of the wavefunction there. In this section, we concentrate on the qualitative features of the equations, because they show us how to unfold the qualitative features of the solutions without the clutter of detail.
2.1 The curvature of the wavefunction
ðd2 f =dx2 Þ
4
2
j
E >V
>0
<0
Fig. 2.4 The variation of the curvature of a wavefunction with the sign of the wavefunction at the point in question and the relative size of the energy and potential energy at the point.
First, we should note that the curvature of c is proportional to the amplitude of c. Therefore, for a given value of V E when c is large, the curvature is large. Where c falls towards zero its curvature decreases (Fig. 2.3). Where c is zero, its curvature is zero. Next, note that where E > V, the factor V E < 0, so the sign of the curvature of c is opposite to the sign of c itself. That is, if E > V and c > 0, then c has negative curvature and looks like . On the other hand, where E < V, V E is positive, and the curvature of c has the same sign as its amplitude. A wavefunction with positive amplitude would then have a positive curvature, and look like . Finally, the curvature is proportional to the difference jV Ej, so if the total energy is greatly in excess of the potential energy (that is, the kinetic energy is high), then the curvature is large. These features are summarized in Fig. 2.4, which contains all the information we need to solve the Schro¨dinger equation qualitatively for a one-particle, one-dimensional system.
2.2 Qualitative solutions Consider a system in which the potential energy depends on position as depicted in Fig. 2.5. Suppose that at x00 the wavefunction has the amplitude and slope as shown as A, and that the total energy of the particle is E. Note that E < V for positions to the right of x 0 but that E > V to the left of x 0 : the sign of E V therefore changes at x 0 . Because cA > 0 at x00 and V < E, the curvature of cA is negative. The wavefunction remains positive at x 0 , but to the right of that point V > E. Its curvature therefore becomes positive, and it bends away from the x-axis and rises to infinity as x increases. Therefore, according to the Born interpretation, c is an inadmissable wavefunction. With this failure in mind, we select a function cB that has a different slope at x00 but the same amplitude. This function has a negative curvature (because E >V). Its curvature becomes positive to the right of x 0 because its amplitude is positive but now E < V. The change in curvature is insufficient to stop cB falling through zero to a negative value, and as it does so its curvature changes sign. This negative curvature forces cB to a negatively infinite value as x increases, and it is therefore an inadmissable wavefunction. Learning from our mistakes, we now select a wavefunction cC that has a slope intermediate between those of cA and cB. Its curvature changes sign at x 0 but it does so in such a way that cC approaches zero asymptotically as x increases. As it does so, its curvature lessens (because the curvature is proportional to the amplitude) and it curls off to neither positive nor negative infinity. Such a wavefunction is acceptable. Note that for the potential shown
j
Potential energy, V
46
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
in Fig. 2.5, a well-behaved wavefunction can be found for any value of E simply by adjusting the amplitude or slope of the function at x00 . Therefore, the energies of such systems are not quantized.
Energy, E
2.3 The emergence of quantization
x Wavefunction,
B
A
C
A C
x
x B
Potential energy, V
Fig. 2.5 The acceptability of a wavefunction is determined by the amplitude and slope at a particular point and the consequent implications on the behaviour of the wavefunction at the boundary. Only C is acceptable.
E E
Wavefunction,
x
C D
Fig. 2.6 When there are two boundary conditions to satisfy (in the sense that the particle is bounded), then it is possible to find acceptable solutions only for certain values of E. That is, the need to satisfy boundary conditions implies the quantization of the energy of the system.
Now that we have seen the sensitivity of the wavefunction to a potential that rises to a large value only on one side, it should be easy to appreciate the difficulty of fitting a function to a system in which the potential confines the particle on both sides (Fig. 2.6). The function cC that was acceptable in the system shown in Fig. 2.5 has been traced to the left, where V rises above E again. We see that its behaviour at this boundary means that cC is unacceptable. In fact, in general it is impossible to find an acceptable solution for an arbitrary value of E. Only for some values of E is it possible to construct a well-behaved function. One such function is cD in Fig. 2.6. In other words, the energy is quantized in a system with a boundary on each side. The considerable importance of this conclusion cannot be overemphasized. The Schro¨dinger equation, being a differential equation, has an infinite number of solutions. It has mathematically acceptable solutions for any value of E. However, the Born interpretation imposes restrictions on the solutions. When the system has boundaries that confine the particle to a finite region, almost all the solutions are unacceptable: acceptable solutions occur only for special values of E. That is, energy quantization is a consequence of boundary conditions. The diagram in Fig. 2.7 depicts the effect of boundaries on the quantization of the energy of a particle. Quantization occurs only when the particle is confined to a finite region of space. When its energy exceeds E 0 the particle can escape to positive values of x, and when its energy exceeds E00 the particle can travel indefinitely to positive and negative values of x. Furthermore, as the potential becomes less confining (that is, when the region for which E > V becomes larger), the separation between neighbouring quantized levels is reduced because it gets progressively easier to find energies that give well-behaved functions. The region of quantized energy is generally taken to signify that we are dealing with bound states of a system, in which the wavefunction is localized in a definite region (like an electron in a hydrogen atom). The region of non-quantized energy is typically associated with scattering problems in which projectiles collide and then travel off to infinity. We introduce both types of solution in this chapter, but delay the complications of scattering problems until Chapter 14 at the end of the book.
2.4 Penetration into non-classical regions A glance at Fig. 2.6 shows that a wavefunction may be non-zero even where E < V; that is, c need not vanish where the kinetic energy is negative. A negative kinetic energy is forbidden classically because v2 cannot be negative, and the fact that a particle may be found in a region where the
2.4 PENETRATION INTO NON-CLASSICAL REGIONS
E0
Energy, E
Unquantized energies E9
Potential energy, V Position, x Fig. 2.7 A general summary of the role of boundaries: the system is quantized only if it is confined to a finite region of space. A single boundary does not entail quantization.
j
47
kinetic energy is negative is an example of quantum mechanical ‘penetration’. We shall elaborate on this term in the course of this chapter. The penetration of a particle into a region where the kinetic energy is negative is no particular cause for alarm. We have seen that observed energies are the expectation values of operators, and the expectation value of the kinetic energy operator is invariably positive (the operator for kinetic energy is proportional to the square of an hermitian operator, px). In addition, because the eigenvalues of the squares of hermitian operators are always positive (Example 1.9), each individual determination of the kinetic energy will have a positive outcome. Finally, any attempt to confine a particle within a nonclassical region, and then to measure its kinetic energy, will be doomed by the uncertainty principle. The confinement would have to be to such a small region that the corresponding uncertainty in momentum, and hence in kinetic energy, would be so great that we would be unable to conclude that the kinetic energy was indeed negative.
Translational motion The easiest type of motion to consider is that of a completely free particle travelling in an unbounded one-dimensional region. Because the potential energy is constant, and may be chosen to be zero, the hamiltonian for the system is H¼
2 d2 h 2m dx2
ð2:3Þ
The time-independent Schro¨dinger equation, Hc ¼ Ec, therefore has the form
2 d2 c h ¼ Ec 2m dx2
The general solutions of this equation are 2mE 1=2 ikx ikx c ¼ Ae þ Be k¼ h2
ð2:4Þ
ð2:5Þ
as may readily be checked by substitution. Because e ikx ¼ cos kx i sin kx (Euler’s relation), an alternative form of this solution is c ¼ C cos kx þ D sin kx
ð2:6Þ
In both forms, the solutions of the coefficients A, B, C, and D are to be found by considering the boundary conditions (see below). However, an important point is that functions of the form e ikx are not square-integrable (Section 1.12), so care needs to be taken with their interpretation. Indeed, because they correspond to a uniform probability distribution throughout space (because je ikxj ¼ 1), they cannot be a description of real physical systems. To cope with this problem we need the concept of wavepacket (Section 2.8).
48
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
2.5 Energy and momentum The first point to note about the solutions is that, as the motion is completely unconfined, the energy of the particle is not quantized. An acceptable solution exists for any value of E: we simply use the appropriate value of k in eqn 2.5. The relation between the energy of a free particle and its linear momentum is E ¼ p2/2m. According to eqn 2.5, the energy is related to the parameter k h2/2m. It follows that the magnitude of the linear momentum of by E ¼ k2 a particle described by the wavefunctions in eqn 2.5 is p ¼ k h
ð2:7Þ
This expression can be developed in a number of ways. For example, we can turn it round, and say that the form of the wavefunction of a particle with linear momentum of magnitude p is given by eqn 2.5 with k ¼ p/ h. A second point is that the wavefunctions in eqn 2.5 have a definite wavelength. This may be easier to see in the case of eqn 2.6, because a wave of wavelength l is commonly written as cos(2px/l) or as sin(2px/l). It follows that the wavelength of the wavefunction in eqn 2.6 is l ¼ 2p/k. That is, the wavefunction for a particle with linear momentum p ¼ k h has a wavelength l ¼ 2p/k. It follows that the wavelength and linear momentum are related by p¼
2p h h¼ l l
ð2:8Þ
This is the de Broglie relation (Section 0.5).
2.6 The significance of the coefficients The significance of the coefficients in the wavefunction can be determined by considering the effect of the linear momentum operator in the position representation, p ¼ ( h/i)d/dx. Suppose initially that B ¼ 0, then ikx h dc h d Ae ¼ pc ¼ ¼ k hAeikx ¼ k hc ð2:9aÞ i dx i dx We see that the wavefunction is an eigenfunction of the linear momentum operator, and that its eigenvalue is k h. Alternatively, if A ¼ 0, then ikx h dc h d Be ¼ ¼ k hBeikx ¼ k hc ð2:9bÞ pc ¼ i dx i dx = eikx
p = – kh
p = + kh
= e –ikx
Fig. 2.8 Wavefunctions for a particle
travelling to the right (towards increasing x) and left (towards decreasing x) with a given magnitude of linear momentum (k h) are each other’s complex conjugate.
The distinction between the two solutions is the sign of the eigenvalue. Because linear momentum is a vector quantity, we are immediately led to the conclusion that the two wavefunctions correspond to states of the particle with the same magnitude of linear momentum but in opposite directions. This is a very important point, for it lets us write down the wavefunctions for particles that not only have a definite kinetic energy and therefore magnitude of linear momentum, but for which we can also specify directions of travel (Fig. 2.8). The significance of the coefficients A and B should now be clearer: they depend on how the state of the particle was prepared. If it was shot from a gun in the direction of positive x, then B ¼ 0. If it had been shot in the opposite
2.7 THE FLUX DENSITY
j
49
direction (by the duelling partner), then its state would be described by a wavefunction with A ¼ 0. Now we turn to the significance of the coefficients C and D in the alternative form of the wavefunction. Suppose D ¼ 0, so that the particle is described by the wavefunction C cos kx. When we examine the effect of the momentum operator we find pc ¼
dc h h dðC cos kxÞ ¼ ¼ ik hC sin kx i dx i dx
We see that the wavefunction is not an eigenfunction of the linear momentum operator. However, by using Euler’s relation and writing c ¼ 12 Ceikx þ 12 Ceikx
Re x Im
(a) Re
x (b)
Im
Fig. 2.9 The relative phases of the imaginary and real components of a wavefunction determine the direction of propagation of the particle: the real component seems to chase the imaginary component. (a) eikx, (b) eikx.
we see that the wavefunction is a superposition of the two linear momentum eigenstates with equal coefficients. From the general considerations set out in Section 1.11, we can conclude that in a series of observations, we would obtain the linear momentum þk h half the time and k h half the time, but we would not be able to predict which direction we would detect in any given observation. The expectation value of the linear momentum, its average value, is zero if its wavefunction is a sine (or a cosine) function. An important general point illustrated by this discussion is that a complex wavefunction (such as eikx), or any function that cannot be made real simply by multiplication by a constant, corresponds to a definite state of linear momentum (in direction as well as in magnitude), whereas a real function (such as cos kx) does not (see Self-test 1.9). To illustrate this point, Fig. 2.9 depicts both the real and imaginary components of a complex wavefunction in a single diagram by plotting the points (cos kx, sin kx) against x. The two functions shown there, (a) eikx and (b) eikx, which correspond to opposite directions of travel, then form two helices, which convey the different senses of motion.
2.7 The flux density Further insight into the form of the general solutions of the Schro¨dinger equation for free particles can be obtained by introducing a quantity called the flux density, Jx. The full usefulness of this quantity will become clear in later chapters where we are interested in the flow of charge in a molecule and the impact of beams of molecules on one another. The flux density in the x-direction is defined as follows: Jx ¼
1 C px C þ Cpx C 2m
ð2:10Þ
In the position representation, we interpret px as ( h/i)d/dx and px ¼ ( h/i) (d/dx). For a state with a definite energy, the time-dependent phase factors in C and C cancel, and the flux density is Jx ¼
1 c px c þ cpx c 2m
ð2:11Þ
50
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
Illustration 2.1 The flux density
To see the significance of the flux density, here we calculate the flux density for a system that is described by the wavefunction in eqn 2.5 with B ¼ 0: 1 ikx h d ikx ikx h d ikx þ Ae Ae Jx ¼ Ae Ae 2m i dx i dx 1 ikx h d ikx ikx h d ikx ¼ Ae A e Ae A e 2m i dx i dx ¼
o k jAj2 n ikx h hjAj2 ðikÞ eikx eikx ðikÞ eikx ¼ e 2mi m
For the wavefunction with A ¼ 0 we find similarly that Jx ¼
khjBj2 m
We now note that k h/m is the classical velocity of the particle, so the flux density is the velocity multiplied by the probability that the particle is in that particular state.
2.8 Wavepackets
Total amplitude,
So far, we have considered a case in which the energy of the particle is specified exactly. But suppose that the particle had been prepared with an imprecisely specified energy. Because the energy is imprecise, the wavefunction that describes the particle must be a superposition of functions corresponding to different energies. Such a superposition is called a wavepacket. For example, suppose the particle is a projectile fired towards positive x; then we know that the wavefunction of the projectile must be a superposition of functions of the form eikx with a range of values of k corresponding to the range of linear momenta (and hence kinetic energies) possessed by the particle. A wavepacket is a wavefunction that has a non-zero amplitude in a small region of space and is close to zero elsewhere. In general, wavepackets move through space in a manner that resembles the motion of a classical particle. To see both these features, we consider a superposition of time-dependent wavefunctions of the form Ck ðx, tÞ ¼ Aeikx eiEk t=h
x
Fig. 2.10 A wavepacket formed by
the superposition of many waves with different wavelengths. Twenty waves have been superimposed to produce this figure.
ð2:12Þ
The superposition is a linear combination of such functions, each one of which is weighted by a coefficient g(k) called the shape function of the packet. Because k is a continuously variable parameter, the sum is actually an integral over k, and so the wavepacket has the form Z ð2:13Þ Cðx, tÞ ¼ gðkÞCk ðx, tÞ dk The pictorial form of such a packet is shown in Fig. 2.10. As a result of the interference between the component waves, at one instant the wavepacket has a large amplitude at one region of space. However, because the time-dependent factor affects the phases of the waves that contribute to the superposition,
2.9 AN INFINITELY THICK POTENTIAL WALL
51
the region of constructive interference changes with time (Fig. 2.11). It should not be hard to believe that the centre of the packet moves to the right, and this is confirmed by a mathematical analysis of the motion (see Further information 5). The classical motion of a projectile is captured by the motion of the wavepacket, and once again we see how classical mechanics emerges from quantum mechanics.
Total amplitude,
j
x
Penetration into and through barriers
Time, t
Fig. 2.11 Because each wave in a
superposition oscillates with a different frequency, the point of constructive interference moves as time increases.
A highly instructive extension of the results for free translational motion is to the case where the potential energy of a particle rises sharply to a high, constant value, perhaps to decline to zero again after a finite distance. Classically we know what happens: if a particle approaches the barrier from the left, then it will pass over it only if its initial energy is greater than the potential energy it possesses when it is inside the barrier. If its energy is lower than the height of the barrier, then the particle is reflected. To see what quantum mechanics predicts, we shall consider three types of barrier of increasing difficulty.
2.9 An infinitely thick potential wall
Potential energy, V (x )
The Schro¨dinger equation for the problem falls apart into two equations, one for each zone in Fig. 2.12. The hamiltonians for the two zones are
V
h 2 d2 2m dx2
Zone I ðx < 0Þ:
H¼
Zone II ðx 0Þ:
h 2 d2 H¼ þV 2m dx2
ð2:14Þ
The corresponding equations are free-particle Schro¨dinger equations, except for the replacement of E by E V in Zone II. Therefore, the general solutions can be written down by referring to eqn 2.5: Zone I
Zone I:
Zone II
k h ¼ f2mEg1=2
c ¼ Aeikx þ Beikx 0
0
Zone II: c ¼ A0 eik x þ B0 eik x 0
Fig. 2.12 The potential energy
of a barrier of finite height but of semi-infinite extent.
x
k0 h ¼ f2mðE VÞg1=2
ð2:15Þ
We shall concentrate on the case when E < V, so that classically the particle cannot be found at x > 0 (inside the wall). The condition E < V implies that k 0 is imaginary; so we shall write k 0 ¼ ik, where k (kappa) is real. It then follows that Zone II: c ¼ A0 ekx þ B0 ekx
k h ¼ f2mðV EÞg1=2
ð2:16Þ
This wavefunction is a mixture of decaying and increasing exponentials: we see that a wavefunction does not oscillate when E < V. Because the barrier is infinitely wide, the increasing exponential must be ruled out because it leads to an infinite amplitude. Therefore, inside a barrier like that shown in Fig. 2.12, the wavefunction must be simply an exponentially decaying function, ekx. One important point about this conclusion is
52
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
that, because the wavefunction is non-zero inside the barrier, the particle may be found inside a classically forbidden region, the effect called penetration. The rapidity with which the wavefunction decays to zero is determined by the value of k, for the amplitude of the wavefunction decreases to 1/e of its value at the edge of the barrier in a distance 1/k, which is called the penetration depth. The penetration depth decreases with increasing mass of the particle and the height of the barrier above the energy of the incident particle (the value of V E). Macroscopic particles have such large masses that their penetration depth is almost zero whatever the height of the barrier, and for all practical purposes they are not found in classically forbidden regions. An electron or a proton, on the other hand, may penetrate into a forbidden zone to an appreciable extent. For example, an electron that has been accelerated through a potential difference of 1.0 V, and which has acquired a kinetic energy of 1.0 eV, incident on a potential barrier equivalent to 2.0 eV, will have a wavefunction that decays to 1/e of its initial amplitude after 0.20 nm, which is comparable to the diameter of one atom. Hence, penetration can have very important effects on processes at surfaces, such as electrodes, and for all events on an atomic scale.
Potential energy, V (x )
2.10 A barrier of finite width We now consider the case of a barrier of a finite width (Fig. 2.13). In particular, the potential energy, V(x), has the form:
V
Zone I ðx < 0Þ: VðxÞ ¼ 0 Zone II ð0 x < LÞ: VðxÞ ¼ V Zone III ðx LÞ: VðxÞ ¼ 0 Zone I
Zone II Zone III 0
L
x
Fig. 2.13 The potential energy
of a finite barrier. Particles incident from one side may be found on the opposite side of the barrier. According to classical mechanics, that is possible only if E is not less than V. According to quantum mechanics, however, barrier penetration may occur whatever the energy.
ð2:17Þ
The general solutions of the time-independent Schro¨dinger equation can be written down immediately: Zone I: Zone II: Zone III:
c ¼ Aeikx þ Beikx 0 0 c ¼ A0 eik x þ B0 eik x c ¼ A00 eikx þ B00 eikx
k h ¼ f2mEg1=2 0 k h ¼ f2mðE VÞg1=2 k h ¼ f2mEg1=2
ð2:18Þ
In scattering problems, of which this is a simple example, it is common to distinguish between ‘incoming’ and ‘outgoing’ waves. An incoming wave is a contribution to the total wavefunction with a component of linear momentum towards the target (from any direction). An outgoing wave is a contribution with a component of linear momentum away from the target. Each contribution corresponds to a flux of particles either towards or away from the target. In the problem we are currently considering, in Zone I A is the coefficient of the incoming wave and B the coefficient of the outgoing wave. In Zone III, A00 is the coefficient of the outgoing wave and B00 the coefficient of the incoming wave. In this section we first consider solutions for E < V. Classically, the particle does not have enough energy to overcome the potential barrier. Therefore, for a particle incident from the left, the probability is exactly zero that it will be found on the right of the barrier (x > L). Quantum mechanically, however, the particle can be found on the right of the barrier even though E < V.
2.10 A BARRIER OF FINITE WIDTH
j
53
In Zone II, the wavefunction has the form given in eqn 2.16. We need to note that the increasing exponential function in the wavefunction in this zone will not rise to infinity before the potential has fallen to zero again and oscillations resume. Therefore, the coefficient B 0 will not be zero. The values of the coefficients are established by using the acceptability criteria for wavefunctions set out at the beginning of this chapter, and in particular the requirement that they and their slopes must be continuous. The continuity condition lets us match the wavefunction at the points where the zones meet, and therefore to find conditions for the coefficients. For example, the continuity of the amplitude at x ¼ 0 and at x ¼ L leads to the two conditions At x ¼ 0: A þ B ¼ A0 þ B0 At x ¼ L: A0 ekL þ B0 ekL ¼ A00 eikL þ B00 eikL
ð2:19Þ
Similarly, the continuity of slopes at the same two points leads to the two conditions At x ¼ 0: ikA ikB ¼ kA0 þ kB0 At x ¼ L: kA0 ekL þ kB0 ekL ¼ ikA00 eikL ikB00 eikL
ð2:20Þ
These four equations give four conditions for finding six unknowns. The remaining conditions include a normalization requirement (one more condition) and a statement about the initial state of the particle (such as the fact that it approaches the barrier from the left). Consider the case where the particles are prepared in Zone I with a linear momentum that carries them to the right. It then follows that the coefficient B00 ¼ 0, because the exponential function it multiplies corresponds to particles with linear momentum towards the left on the right-hand side of the barrier, and there can be no such particles. That is, there is no incoming wave, no inward flux of particles, in Zone III. There may be particles travelling to the left on the left of the barrier because reflection can take place at the barrier. We can therefore identify the coefficient B as determining (via jBj2) the flux density of particles reflected from the barrier in Zone I. The reflection probability, R, is the ratio of the reflected flux density to the incident flux density, so from the results of Illustration 2.1 we can write (disregarding signs): R¼
ðk h=mÞjBj2 ðk h=mÞjAj2
¼
jBj2
ð2:21aÞ
jAj2
Similarly, the coefficient A00 , the coefficient of the outgoing wave in Zone III, determines (via jA00 j2) the flux of particles streaming away from the barrier on the right. The transmission probability, T, is the ratio of the transmitted flux density to the incident flux density, and is given by T¼
jA00 j2
ð2:21bÞ
jAj2
The complete calculation of T involves only elementary manipulations of the relations given above, and the result is T¼
1 1þ
ðekL
2 ekL Þ =f16ðE=VÞð1
E=VÞg
R¼1T
ð2:22Þ
54
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
Transmission probability, T
0.5
0.4
2
0.3
0.2
3 4
0.1
6 8 10
0
0
0.2
0.4
E/V
0.6
0.8
1.0
Fig. 2.14 The tunnelling probability
through a finite rectangular barrier as a function of incident energy. The curves are labelled with the value of Lð2mVÞ1=2 = h.
To obtain this result, we have used the first of the two relations eix eix sin x ¼ 2i eix þ eix cos x ¼ 2
We shall use the fact a number of times that sin x ¼ 0 at x ¼ np with n ¼ 0, 1, 2, . . . , and cos x ¼ 0 at x ¼ np/2, with n ¼ 1, 3, 5, . . . .
with k ¼ {2mV(1 E/V)}1/2/h. Because we have been considering energies E < V, T represents the probability that a particle incident on one side of the barrier will penetrate the barrier and emerge on the opposite side. That is, T is the probability of tunnelling, non-classical penetration, through the barrier (Fig. 2.14). We now deal with energies E > V. Classically, the particle now has sufficient energy to overcome the potential barrier. A particle incident from the left would have unit probability of being found on the right of the barrier. Once again, though, quantum mechanics gives a different result. To determine the expressions for T and R we could proceed as we did above for energies E < V, write down four relations for the six coefficients, and then manipulate them. However, it is considerably easier to take the expression for T given above and replace k by k 0 /i ¼ ik 0 . This procedure gives T¼
1 1 þ ðsin2 ðk0 LÞÞ=f4ðE=VÞðE=V 1Þg
R¼1T
ð2:23Þ
with hk0 ¼ f2mVðE=V 1Þg1=2 . This function is plotted in Fig. 2.15. The transmission coefficient, T, takes on its maximum value of 1 and the barrier is transparent when sin(k 0 L) ¼ 0, which occurs at energies E corresponding to1 np n ¼ 1, 2, . . . ð2:24aÞ k0 ¼ L Furthermore, T has minima near np k0 ¼ n ¼ 1, 3, . . . 2L
ð2:24bÞ
At high energies (E V), T approaches its classical value of 1. We see in Fig. 2.15 how the transmission coefficient for energies above the barrier height fluctuates between maxima and minima. We should take note of two striking differences between the quantum mechanical and classical results. First, even when E > V, there is still a probability of the particle being reflected by the potential barrier even though classically it has enough energy to travel over the barrier. This phenomenon is known as antitunnelling or non-classical reflection. Second, the strong variation of T with the energy of the incident particle is a purely quantum mechanical effect. The peaks in the transmission coefficient for energies above V are examples of scattering resonances. We shall have more to say concerning resonances in Chapter 14 when we discuss scattering in general.
2.11 The Eckart potential barrier The rectangular barrier we have been considering is obviously not very realistic, but it does serve to introduce a number of concepts, and it has properties that are found in more realistic models. In fact, there are only a few .......................................................................................................
1. The value n ¼ 0 is excluded because in the limit of k 0 ! 0, T ¼ 1/(1 þ mVL2/2 h2), which is not equal to 1.
2.11 THE ECKART POTENTIAL BARRIER
Transmission probability, T
2 0.8
VðxÞ ¼ 0.6 3 0.4
0.2 10
2
55
realistic potentials for which analytical expressions for the reflection and transmission coefficients are available. One such potential is the Eckart potential barrier:2
1
0 1
j
E /V
3
4
Fig. 2.15 The same as in the
preceding illustration, but for E > V. Note that according to quantum mechanics, the particle may be reflected back from the barrier (so that P < 1) even though classically it has enough energy to pass over it.
The hyperbolic sin (sinh) and cosine (cosh) functions are defined as sinh x ¼
ex ex 2
cosh x ¼
ex þ ex 2
4V0 ebx ð1 þ ebx Þ
2
ð2:25Þ
where V0 and b are constants with dimensions of energy and inverse length, respectively. The potential is shown in Fig. 2.16; we see that it is symmetric in x with a maximum value of V0 at x ¼ 0, and approaches zero as jxj ! 1. The Schro¨dinger equation associated with this potential can be solved, but its solutions are the so-called hypergeometric functions, which are beyond the scope of this book. All we shall do is quote the analytical expression for the transmission coefficient: n o cosh 4pð2mEÞ1=2 = hb 1 h T¼ i1=2 n o hb þ cosh 2p 8mV0 ð hb=2Þ2 = hb cosh 4pð2mEÞ1=2 = ð2:26Þ Figure 2.17 shows how T varies with E/V0. For E V0, T 0. As the energy approaches the top of the barrier (E ¼ V0), the transmission probability increases. This increase corresponds to the tunnelling of the particle through the barrier and its emergence on the other side. As the energy increases beyond V0, T approaches 1, but T < 1 even when E > V0. There is still a probability of the particle being reflected by the barrier even when classically it can pass over it. This behaviour is another example of the antitunnelling displayed by the rectangular barrier. Finally, when E V0, T 1, as expected classically. However, unlike the rectangular barrier, there are no oscillations in the transmission probability for E > V0.
1
0.8
Particle in a box We now turn to a case in which a particle is confined by walls to a region of space of length L. The walls are represented by a potential energy that is zero inside the region but rises abruptly to infinity at the edges (Fig. 2.18). This system is called a one-dimensional square well or a particle in a box. The squareness in the former name refers to the steepness with which the potential energy goes to infinity at the ends of the box. Because the particle is confined, its energy is quantized, and the boundary conditions determine which energies are permitted.
V (x )/V0
0.6
0.4
0.2
0 –6
–4
–2
0 x
2
4
6 .......................................................................................................
Fig. 2.16 The Eckart potential
barrier, as described in the text.
2. This barrier was investigated by C. Eckart in 1930. For details, see C. Eckart, Phys. Rev., 1303, 35 (1930).
56
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR 1
Transmission probability, T
2.12 The solutions 0.8
The hamiltonian for the system is
0.6
H¼
0
0
1 2
0.5
1
1.5 E /V0
VðxÞ ¼
0 1
for 0 x L otherwise
ð2:27Þ
Because the potential energy of a particle that touches the walls is infinite, the particle cannot in fact penetrate them. This result is justified by the behaviour of the wavefunctions described in Section 2.9. It follows that the hamiltonian for the region where the potential is not infinite, and therefore the only region where the wavefunction is non-zero, is
0.4
0.2
2 d2 h þ VðxÞ 2m dx2
2
2.5
H¼
Fig. 2.17 The tunnelling probability
Potential energy, V (x )
for an Eckart barrier and its variation with energy. The curves are labelled with the value of ð2mV0 Þ1=2 =bh.
2 d2 h 2m dx2
ð2:28Þ
This expression is the same as the hamiltonian for free translational motion (eqn 2.3), so we know at once that the solutions are those given in eqn 2.6. However, in this case there are boundary conditions to satisfy, and they will have the effect of eliminating most of the possible solutions. The wavefunctions are zero outside the box where x < 0 or x > L. Wavefunctions are everywhere continuous. Therefore, the wavefunctions must be zero at the walls at x ¼ 0 and x ¼ L. The boundary conditions are therefore c(0) ¼ 0 and c(L) ¼ 0. We now apply each condition in turn to a general solution of the form cðxÞ ¼ C cos kx þ D sin kx
k h ¼ ð2mEÞ1=2
First, at x ¼ 0, cð0Þ ¼ C cos 0 þ D sin 0 ¼ C because cos 0 ¼ 1 and sin 0 ¼ 0. Therefore, to satisfy the condition c(0) ¼ 0 we require C ¼ 0. Next, at x ¼ L, after setting C ¼ 0, cðLÞ ¼ D sin kL
Fig. 2.18 The infinite square-well
potential characteristic of a particle in a box.
One way to achieve c(L) ¼ 0 is to set D ¼ 0, but then the wavefunction would be zero everywhere and the particle found nowhere. The alternative is to require that the sine function itself vanishes. It does so if kL is equal to an integral multiple of p. That is, we must require k to take the values k¼
np L
n ¼ 1, 2, . . .
ð2:29Þ
The value n ¼ 0 is excluded because it would give sin kx ¼ 0 for all x, and the particle would not be found anywhere. The integer n is an example of a quantum number, a number that labels a state of the system and that, by the use of an appropriate formula, can be used to calculate the value of an h2/2m, it follows that observable of the system. For instance, because E ¼ k2 the energy is related to n by En ¼
n2 h 2 p2 n2 h2 ¼ 2 2mL 8mL2
n ¼ 1, 2, . . .
ð2:30Þ
2.13 FEATURES OF THE SOLUTIONS
To evaluate this integral we have used the standard form Z sin2 ax dx ¼ 12x
1 sin 2ax þ constant 4a
Potential energy, V (x )
57
A major conclusion of this calculation at this stage is that the energy of the particle is quantized; that is, confined to a series of discrete values. There now remains only the constant D to determine before the solution is complete. The probability of finding the particle somewhere within the box must be 1, so the integral of c2 over the region between x ¼ 0 and x ¼ L must be equal to 1. The integral is Z L Z L npx c c dx ¼ D2 sin2 dx ¼ 12LD2 L 0 0 Therefore, as we saw in Example 1.4, D ¼ (2/L)1/2. The complete solution is 1=2 2 npx sin cðxÞ ¼ L L ð2:31Þ 2 2 n h n ¼ 1, 2, . . . En ¼ 8mL2
6
5
4 3 2 1 L
0
x
Fig. 2.19 The first six energy levels
and the corresponding wavefunctions for a particle in a box. Notice that the levels are more widely separated as the energy increases; the maximum amplitude of the wavefunctions is the same in all cases.
6
Potential energy, V (x )
j
We see that there is a single quantum number, n, which determines the wavefunctions and the energies. Figure 2.19 shows some of the solutions and Fig. 2.20 shows the squares of the wavefunctions: the latter are the probability densities for finding the particle in each location. Note how the particle seems to avoid the walls in the low energy states but becomes increasingly uniformly distributed as n increases. The distribution at high values of n corresponds to the classical expectation that the particle spends, on the average, equal times at all points as it bounces between the walls. This behaviour is an example of the correspondence principle, which states that classical mechanics emerges from quantum mechanics at high quantum numbers. A point where a wavefunction passes through zero (not simply approaches zero without passing through) is called a node. We see from Fig. 2.19 that the lowest energy state has no nodes, and that the number increases as n increases: in general, the number of nodes is n 1. It is a common feature of wavefunctions that the higher the number of nodes, the higher the energy. With more nodes, there is greater curvature of the wavefunction and therefore a greater kinetic energy.
2.13 Features of the solutions 5 4 3 2 1 0
L
x
Fig. 2.20 The probability
distribution of a particle in a box. Note that the distribution becomes more uniform as the energy increases.
The lowest energy that the particle can have is for the state with n ¼ 1, its lowest value, and is E1 ¼ h2/8mL2. This irremovable energy is called the zero-point energy. It is a purely quantum mechanical property, and in a hypothetical universe in which h ¼ 0 there would be no zero-point energy. The uncertainty principle gives some insight into its origin, because the uncertainty in the position of the particle is finite (it is somewhere between 0 and L), so the uncertainty in the momentum of the particle cannot be zero. Because Dp 6¼ 0, it follows that hp2i 6¼ 0 and consequently that the average kinetic energy, which is proportional to hp2i, also cannot be zero. A more fundamental way of understanding the origin of the zero-point energy, though, is to note that the wavefunction is necessarily curved if it is to be zero at each wall but not zero throughout the interior of the box. We have already
58
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
seen that the curvature of a wavefunction signifies the possession of kinetic energy, so the particle necessarily possesses non-zero kinetic energy if it is inside the box. The energy separation of neighbouring states decreases as the walls move back and give the particle more freedom: n o h2 h2 ¼ ð2n þ 1Þ Enþ1 En ¼ ðn þ 1Þ2 n2 2 8mL 8mL2
ð2:32Þ
As the length of the box approaches infinity (corresponding to a box of macroscopic dimensions), the separation of neighbouring levels approaches zero, and the effects of quantization become completely negligible. In effect, the particle becomes unbounded and free, and its state is described by the wavefunctions in eqn 2.5. The same is true as the mass, m, becomes large. Consequently, classical mechanics can be used to describe the translational motion of macroscopic objects.
2.14 The two-dimensional square well V (x, y )
L2 y 0
x
L1
Fig. 2.21 An exploded view of the
potential energy of a particle in a two-dimensional square well.
Interesting new features arise when we consider a particle confined to a rectangular planar surface with linear dimensions L1 in the x direction and L2 in the y direction (Fig. 2.21). Just as in one dimension, where the wavefunctions look like those of a vibrating string with clamped ends, so in two dimensions they can be expected to correspond to the vibrations of a plate with the edges rigidly clamped. The hamiltonian for the two-dimensional, infinitely deep square well in the interior of the well (the only region where the particle will be found, and where its potential energy is zero) is ! h2 q2 q2 ð2:33Þ þ H¼ 2m qx2 qy2 The Schro¨dinger equation for the particle inside the walls is therefore q2 c q2 c 2mE þ ¼ 2 c qx2 qy2 h
ð2:34Þ
The boundary conditions are that the wavefunction must vanish at all four walls. To solve this equation in two variables, we try the separation of variables technique described in Section 1.14. The trial solution is written c(x,y) ¼ XY, where X is a function of only x and Y is a function only of y. Inserting the trial solution into the Schro¨dinger equation we get first Y
d2 X d2 Y 2mE þ X ¼ 2 XY dx2 dy2 h
and then, after dividing through by XY, 1 d2 X 1 d2 Y 2mE þ ¼ 2 X dx2 Y dy2 h
2.15 DEGENERACY
j
59
We now use the same argument as in Section 1.14, and conclude that the original equation can be separated into two parts: (a)
(b)
(c)
Fig. 2.22 Three wavefunctions for a
particle in a two-dimensional square well: (a) n1 ¼ 1, n2 ¼ 1, (b) n1 ¼ 2, n2 ¼ 1, and (c) n1 ¼ 2, n2 ¼ 2.
(a)
d2 X 2mEX ¼ X 2 dx h2
d2 Y 2mEY ¼ Y 2 dy h2
with EX þ EY ¼ E. Both equations have the same form as the equation for a one-dimensional system, and the boundary conditions are the same. Therefore, we may write the solutions immediately (using c ¼ XY): 2 n1 px n2 py sin sin cn1 n2 ðx; yÞ ¼ L1 L2 ðL1 L2 Þ1=2 ð2:35Þ h2 n21 n22 n1 ¼ 1, 2, . . . n2 ¼ 1, 2, . . . þ En1 n2 ¼ 8m L21 L22 Note that to define the state of a particle in a two-dimensional system, we need to specify the values of two quantum numbers; n1 and n2 can take any integer values in their range independently of each other. Many of the features of the one-dimensional system are reproduced in higher dimensions. There is a zero-point energy (E1,1), and the energy separations decrease as the walls move apart and become less confining. The energy is quantized as a consequence of the boundary conditions. The shapes of some of the low-energy wavefunctions are illustrated in Fig. 2.22 and the corresponding probability densities are shown in Fig. 2.23. As in the onedimensional case, the particle is distributed more uniformly at high energies than at low.
2.15 Degeneracy One feature found in two dimensions but not in one dimension is apparent when the box is geometrically square. Then L1 ¼ L2 ¼ L and the energies are given by En1 n2 ¼
(b)
(c)
Fig. 2.23 Three probability
distributions for a particle in a two-dimensional square well: (a) n1 ¼ 1, n2 ¼ 1, (b) n1 ¼ 2, n2 ¼ 1, and (c) n1 ¼ 2, n2 ¼ 2 (as in the previous illustration).
h2 2 n þ n22 8mL2 1
ð2:36Þ
This expression implies that a state with the quantum numbers n1 ¼ a and n2 ¼ b (which we could denote ja,bi) has exactly the same energy as one with n1 ¼ b and n2 ¼ a (the state jb,ai) even when a 6¼ b. This is an example of the degeneracy of states mentioned in Section 1.2. For example, the two states j1,2i and j2,1i both have the energy 5h2/8mL2 but their two wavefunctions are different: 2 px 2py 2 2px py c2;1 ðx; yÞ ¼ sin sin sin c1;2 ðx; yÞ ¼ sin L L L L L L Inspection of Fig. 2.24 shows the origin of this degeneracy: one wavefunction can be transformed into the other by rotation of the box through 90 . We should always expect degeneracies to be present in systems that have a high degree of symmetry, as we shall see in more detail in Chapter 5.
j
60
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
– +
(b)
x
y + –
(a) Fig. 2.24 A contour representation of the two degenerate states (a) n1 ¼ 2, n2 ¼ 1 and (b) n1 ¼ 1, n2 ¼ 2 for a particle in a square square well. Note that one wavefunction is rotated into the other by a symmetry transformation of the box (its rotation through 90 about a vertical axis). In this perspective view, the plane looks oblong; it is in fact square.
In the case of a rectangular but not square box, the symmetry and the degeneracy are lost. However, sometimes degeneracy is encountered where there is no rotation that transforms one wavefunction into another; it is then called accidental degeneracy. In certain cases, accidental degeneracy is known to arise when the full symmetry of the system has not been recognized, and a deeper analysis of the system shows the presence of a hidden symmetry that does interrelate the degenerate functions. It may be the case that all accidental degeneracies can be traced to the existence of hidden symmetries. Accidental degeneracy occurs in the hydrogen atom, and we shall continue the discussion there. Example 2.1 Hidden symmetry and accidental degeneracy
Show that in a rectangular box with sides L1 ¼ L and L2 ¼ 2L there is an accidental degeneracy between the states j1,4i and j2,2i. Method. To confirm the degeneracy, all we need do is to substitute the data
into the expression for the energy, eqn 2.35. Answer. The two states have the following energies:
E1;4
h2 12 42 ¼ þ 8m L2 ð2LÞ2
E2;2
h2 22 22 ¼ þ 2 8m L ð2LÞ2
! ¼
5h2 8mL2
¼
5h2 8mL2
!
The energies are the same, despite the lack of symmetry. Comment. In fact, inspection of the wavefunctions (Fig. 2.25) shows that there
is a kind of hidden symmetry, as one half of the box can be rotated relative to the other half, and as a result the two wavefunctions are interconverted, including their behaviour at their nodes and at the walls.
(a) x y
Self-test 2.1. Find other examples of degeneracy in this system. [For instance, the pair (j2,8i, j4,4i)]
The harmonic oscillator (b) Fig. 2.25 An example of accidental
degeneracy: the two functions shown here schematically are degenerate even though one cannot be transformed into the other by a symmetry transformation of the system. Note, however, that a hidden symmetry (the separate rotation of the two halves of the box) does interconvert them.
We now turn to one of the most important individual topics in quantum mechanics, the harmonic oscillator. Harmonic oscillations occur when a system contains a part that experiences a restoring force proportional to the displacement from equilibrium. Pendulums and vibrating strings are familiar examples. An example of chemical importance is the vibration of atoms in a molecule. Another example is the electromagnetic field, which can be treated as a collection of harmonic oscillators, one for each frequency of radiation present. The importance of the harmonic oscillator also lies in the way that the same algebra occurs in a variety of different problems; for example, it also occurs in the treatment of rotational motion.
2.16 THE SOLUTIONS
Energy, E/h
Potential energy, V
10 9 8 7 6 5
4.5 3.5 2.5
4 3 2
1.5
1
0.5 0
0
0 Displacement, x
Fig. 2.26 The parabolic potential
energy characteristic of a harmonic oscillator and the evenly spaced ladder of allowed energies (which continues up to infinity).
j
61
The restoring force in a one-dimensional harmonic oscillator is given by Hooke’s law as kx, where the constant of proportionality k is called the force constant. A stiff spring has a large force constant; a weak spring has a small one. Because the force acting on a particle is the negative gradient of the potential energy (F ¼ dV/dx), it follows that the potential energy of the oscillator varies with displacement x from equilibrium as VðxÞ ¼ 12 kx2
ð2:37Þ
and a graph of potential energy against displacement is a parabola (Fig. 2.26). The difference between this potential and the square-well potential is the rapidity with which it rises to infinity: the ‘walls’ of the oscillator are much softer, and so we should expect the wavefunctions of the oscillator to penetrate them slightly. In other respects the two potentials are similar, and we can imagine the slow deformation of the square well into the smooth parabola of the oscillator. The wavefunctions of one system should change slowly into those of the other: they will have the same general form, but will penetrate into classically disallowed displacements. Another point about the harmonic oscillator is that it is really much too simple. Its simplicity arises from the symmetrical occurrence of momentum and displacement in the expression for the total energy. Classically, the energy is E ¼ p2/2m þ kx2/2, and both p and x occur as their squares. This hidden symmetry has important implications, one being that if there is a new theory that can be applied to the harmonic oscillator and solved, then it may still be unsolvable for other systems. Another implication involves the uncertainty principle, for in the ground state of the harmonic oscillator, the product of the h (see Problem 2.29). uncertainties Dp and Dx is equal to 12
2.16 The solutions Because the potential energy is V ¼ 12 kx2 , the hamiltonian operator for the harmonic oscillator of mass m and force constant k is H¼
2 d2 h þ 1kx2 2m dx2 2
ð2:38Þ
The Schro¨dinger equation is therefore
2 d2 c 1 2 h þ kx c ¼ Ec 2m dx2 2
ð2:39Þ
The best method for solving this equation—a method that also works for rotational motion and the hydrogen atom—is set out in Further information 6. This method depends on looking for a way of factorizing the hamiltonian and introduces the concepts of ‘creation and annihilation operators’. The conventional solution, which involves expressing the solutions as polynomials in the displacement, is described in Further information 7. That algebra, however, need not deflect us from the main thread of this chapter, the discussions of the solutions themselves. As might be expected for such a highly symmetrical system, their properties are remarkably simple.
j
62
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
Factorial n, denoted n!, is the product n! ¼ n(n 1)(n 2) . . . 1, with 0! ¼ 1 by definition. The factorials of large values of n can be estimated from Stirling’s approximation, n! ð2pÞ1=2 nnþ1=2 en 1.0
2 0
4
Wavefunction,
0.5
–0.5
(a)
–4
0 x
–2
1.0 3
2
4
1
Wavefunction,
0
–0.5
(b)
When v ¼ 1, because H1(z) ¼ 2z, the wavefunction is the same gaussian function multiplied by 2ax, with a different normalization factor: 3 1=2 a 1=2 2a 2 2 a2 x2 =2 c1 ðxÞ ¼ 2axe ¼ xea x =2 2p1=2 p1=2 Table 2.1 Hermite polynomials
0.5
–1.0
These energy levels are illustrated in Fig. 2.26. The wavefunctions are no longer the simple sine functions of the square well, but do show a family resemblance to them. They can be pictured as sine waves that collapse towards zero at large displacements (Fig. 2.27). Their precise form is that of 2 a bell-shaped gaussian function, a function of the form ex , multiplied by a polynomial in the displacement: 1=4 mk 2 2 a¼ cv ðxÞ ¼ Nv Hv ðaxÞea x =2 h2 1=2 a Nv ¼ ð2:41Þ 2v v!p1=2 The parameter a has the dimensions of 1/length (so ax is dimensionless). The Hv(z) are Hermite polynomials (Table 2.1). Because H0(z) ¼ 1, the wavefunction for the state with v ¼ 0 is proportional to the bell-shaped 2 2 gaussian function ea x =2 : a 1=2 2 2 ea x =2 c0 ðxÞ ¼ 1=2 p
0
–1.0
The energy of a harmonic oscillator is quantized (as expected from the shape of the potential) and limited to the values 1=2 k ho where o ¼ v ¼ 0; 1; 2; . . . ð2:40Þ Ev ¼ ðv þ 12Þ m
–4
–2
0 x
2
4
Fig. 2.27 The wavefunctions of
a harmonic oscillator for v up to 4: (a) even values, (b) odd values. Note that the number of nodes increases with v, and that even v functions are symmetric whereas odd v functions are antisymmetric about x ¼ 0.
v
Hv (z)
0
1
1
2z
2
4z2 2
3
8z3 12z
4
16z4 48z2 þ 12
5
32z5 160z3 þ 120z
6
64z6 480z4 þ 720z2 120
7
128z7 1344z5 þ 3360z3 1680z
8
256z8 3584z6 þ 13440z4 13440z2 þ 1680
Differential equation: Hv00 2zHv0 þ 2vHv ¼ 0 Recursion relation: R 1 Hv þ 1 ¼ 2zHv 22vHv 1 Orthogonality: R1 Hv ðzÞHv0 ðzÞez dz ¼ 0 for v 6¼ v0 2 1 Normalization: 1 Hv ðzÞ2 ez dz ¼ p1=2 2v v!
2.17 PROPERTIES OF THE SOLUTIONS
j
63
Example 2.2 The nodes of harmonic oscillator wavefunctions
Locate the nodes of the harmonic oscillator wavefunction with v ¼ 4. Method. The gaussian function has no nodes, so we need to determine the
nodes of the Hermite polynomials by determining the values of x at which they pass through zero. The polynomials are listed in Table 2.1. We will need the solutions of a quadratic equation: 1=2 b b2 4ac ax2 þ bx þ c ¼ 0 x¼ 2a Answer. Because H4(ax) ¼ 16(ax)4 48(ax)2 þ 12, we need to solve
16ðaxÞ4 48ðaxÞ2 þ 12 ¼ 0 v 12 11 10 9 8 7 6 5 4 3 2 1 0
–3 –2 –1 0 +1 +2 +3
x
Fig. 2.28 The distribution of
nodes in the first 13 states of a harmonic oscillator (up to v ¼ 12). The white regions show where the wavefunction is positive and the shaded regions where it is negative.
This is a quadratic equation in z ¼ (ax)2, 16z2 48z þ 12 ¼ 0 with roots z¼
1=2 48 482 4 16 12 ¼ 2:7247 and 0:2753 2 16
The nodes are therefore at x ¼ 1.6507/a and 0.5246/a (see Fig. 2.27). Comment. For more complicated polynomials it is best and sometimes
essential to use numerical methods (the root extracting program of a mathematics package). The graph in Fig. 2.28 shows the pattern of nodes: note how they spread away from the origin but become more uniformly distributed as v increases. Self-test 2.2. Identify the location of the five nodes of H5. [At ax ¼ 0, 0.959, 2.020]
2.17 Properties of the solutions Table 2.2 summarizes the properties of the harmonic oscillator. The most significant point about the energy levels is that they form a ladder with equal spacing. The energy separation between neighbours is Evþ1 Ev ¼ ho
ð2:42Þ
regardless of the value of v. The equal spacing of the energy levels is another consequence of the hidden x2, p2 symmetry of the harmonic oscillator. As the force constant k increases, so the separation between neighbouring levels also increases (o / k1/2). As k decreases or the mass increases, so o decreases, and the separation between neighbouring levels decreases too. In the limit of zero force constant the parabolic potential fails to confine the particle (it corresponds to an infinitely weak spring) and the energy can vary continuously. There is no quantization in this limit of an unconstrained, free particle. When thinking about the contributions to the total energy of a harmonic oscillator we have to take into account both the kinetic energy
64
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
Table 2.2 Properties of the harmonic oscillator
Energies:
Ev ¼ (v þ 12) ho, o ¼ (k/m)1/2
Wavefunctions:
cv ðxÞ ¼ Nv H v ðaxÞea x =2 1=4 1=2 mk a a¼ N ¼ v 2v v!p1=2 h2
2 2
h 1=2 h 1=2 1=2 ðv þ 1Þ1=2 hv 1jxjvi ¼ v 2mo 2mo hmo 1=2 hmo 1=2 1=2 hv þ 1jpx jvi ¼ i ðv þ 1Þ1=2 hv 1jpx jvi ¼ i v 2 2
Matrix elements: hv þ 1jxjvi ¼
Virial theorem:
hEK i ¼ hEP i
for all v
(which depends on the curvature of the wavefunction) and the potential energy (which depends on the probability of the particle being found at large displacements from equilibrium). The discussion of the balance between the kinetic and potential contributions to the total energy is greatly simplified by the virial theorem, which although originally derived from classical mechanics has a quantum mechanical counterpart (see Further information 3). The virial theorem states that If the potential energy can be expressed in the form V ¼ axs, where a is a constant, then the mean kinetic, EK, and potential, EP, energies are related by 2hEK i ¼ shEP i
ð2:43Þ
It follows that the total mean energy is E ¼ hEK i þ hEP i ¼ ð1 þ 2=sÞhEK i
ð2:44Þ
For the harmonic oscillator, s ¼ 2, so hEKi ¼ hEPi (hidden symmetry again), and therefore E ¼ 2hEKi. Consequently, as the total energy increases (as it does as k increases for a given quantum state), both the kinetic and the potential energy increase. Not only does the curvature of the wavefunction increase, but the wavefunction also spreads into regions of higher potential energy. In classical terms, this behaviour corresponds to a pendulum swinging more rapidly and with greater amplitude as its energy is increased. ho: The A harmonic oscillator has a zero-point energy of magnitude E0 ¼ 12 classical interpretation of such a conclusion is that the oscillator never stops fluctuating about its equilibrium position. The reason for the existence of the zero-point energy is the same as for a particle in a box: the wavefunctions must be zero at large displacements in either direction (because the potential energy is confining), non-zero in between (because the particle must be somewhere), and continuous (as for all wavefunctions). These conditions can be satisfied only if the wavefunction has curvature; hence the expectation value of the kinetic energy of the oscillator must be non-zero in all its states. By the virial theorem, the expectation values of the kinetic and potential energies are equal in each state, therefore the expectation value of the energy
2.18 THE CLASSICAL LIMIT
j
65
is non-zero even in its lowest state. This argument can also be turned round: if E ¼ 0, then for an oscillator hEKi ¼ hEPi ¼ 0, which implies that both hp2i ¼ 0 and hx2i ¼ 0. For these to be possible mean values, both p and x must be zero, which is contrary to the uncertainty principle.
Probability distribution
2.18 The classical limit
Classical distribution
v = 12
Displacement Fig. 2.29 A comparison of the
probability distribution for a highly excited state of a harmonic oscillator (v ¼ 12) and that of a classical oscillator with the same energy. Note how the former is starting to resemble the latter.
The shapes of the wavefunctions have already been drawn in Fig. 2.27. Their similarities to the square-well wavefunctions should be noted. The major difference between the two is the penetration of the harmonic oscillator wavefunctions into classically forbidden regions where E < V. In the same way as for the square well, the particle clusters away from the walls (and stays close to x ¼ 0) in its lowest energy states. This is the behaviour to be expected classically of a stationary particle, for such a particle will be found at zero displacement and nowhere else. When the oscillator is moving, the classical prediction is that it has the highest probability of being found at its maximum displacement, the turning points of its classical trajectory, where it is briefly stationary. The behaviour of the quantum oscillator is quite different for low energy levels, but the two descriptions become increasingly similar as it is excited into higher levels. We see from Fig. 2.29 that at high v, the wavefunctions have their dominant maxima close to the classical turning points and resemble the classical distribution, as we would expect from the correspondence principle. When the energy levels of the oscillator are close in comparison with the precision with which its state can be prepared (for example, when the parabolic potential is so broad or the mass so great that the levels lie close together), the state of the oscillator must be expressed as a superposition of the wavefunctions considered so far. For example, because the energy levels are only about 1034 J apart for a pendulum of period 1 s, we cannot hope to set it swinging with such precision that we can be confident that only one level is occupied. Setting the pendulum swinging results in its being described by a superposition of wavefunctions, and the interference between the components of the superposition results in the formation of a wavepacket. The timedependence of the components results in a region of constructive interference that moves from one side of the potential to the other with an angular frequency o. That is, for coarse preparations of initial states, there is a sharply defined wavepacket which oscillates in the potential with the angular frequency o ¼ (k/m)1/2. This is precisely the classical behaviour of an oscillator, with the wavepacket denoting the location of the classical particle. In other words, when we see a pendulum swing, we are seeing a display of the separation of its quantized energy levels. Example 2.3 The construction and motion of a wavepacket
Show that whatever superposition of harmonic oscillator states is used to construct a wavepacket, it is localized at the same place at the times 0, T, 2T, . . . , where T is the classical period of the oscillator.
66
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
Method. The classical period is T ¼ 2p/o. We need to form a time-dependent
wavepacket by superimposing the Cv(x,t) for the oscillator, and then evaluate it at t ¼ nT, with n ¼ 0, 1, 2, . . . . Answer. The wavepacket has the following form:
Cðx; tÞ ¼
X
cv Cv ðx; tÞ ¼
cv cv ðxÞe v iðvþ1=2Þot
v
¼
X
X
iEv t= h
cv cv ðxÞe
v
It follows that Cðx; nTÞ ¼
X
cv cv ðxÞe2npiðvþ1=2Þ ¼
v
X
cv cv ðxÞð1Þn
v
¼ð1Þn Cðx; 0Þ because e2pi ¼ 1 and eip ¼ 1. Comment. The wavefunction changes sign after each period T, but is otherwise
unchanged. Because the probability density is proportional to the square of the amplitude, it follows that the original distribution of the particle is recovered after each successive period (Fig. 2.30). Self-test 2.3. Construct the explicit form of C at x ¼ 0 and discuss its time
behaviour.
Translation revisited: The scattering matrix The properties of matrices are reviewed in Further information 23; in this section we deal only with 2 2 matrices, and the manipulations required are very straightforward. Only matrix multiplication is required: a b x ax þ by ¼ dx þ cy d c y a b w x d c z y aw þ bz ax þ by ¼ dw þ cz dx þ cy
In this concluding brief section of the chapter we return to the discussion of unbound translational motion and show that it can be expressed more succinctly. The aim of this section is to introduce one of the most important concepts in scattering theory, the scattering matrix. To do so, we shall redevelop the finite barrier problem treated in Section 2.10, and express it in a way that utilizes this concept. The material here will be developed further in Chapter 14 and could be ignored at this stage. We pick up the finite-barrier story at eqn 2.19 and express the relations between the coefficients in forms of matrices. The coefficients will be written as follows: 0 00 A A A C00 ¼ C¼ C0 ¼ 0 B B00 B for Zones I, II, and III, respectively. The two equations relating the coefficients A, B, A 0 , and B 0 for the wavefunction in Zones I and II can now be expressed in matrix form as: 0 1 1 ik=k 1 þ ik=k C ¼ MC M¼2 ð2:45Þ 1 þ ik=k 1 ik=k
TRANSLATION REVISITED: THE SCATTERING MATRIX
(a)
–4
We have the connection between Zone I and Zone II and between Zone II and Zone III in matrix form. The connection between the coefficients in Zones III and I is now easy to deduce by combining the two relations: C00 ¼ TC
0 –2 2 Displacement, x
(b)
4
T ¼ QM
Probability density, P (x, t )
ð2:47Þ
In exactly the same way, we can set up a matrix relation between the coefficients of the outgoing and incoming waves. First we write 00 00 B A Cin ¼ Cout ¼ ð2:48Þ A B Then the two are related by
2
Cout ¼ SCin 1
ð2:49Þ
Some straightforward algebra shows that the matrices S and T are related by T21 =T22 T11 T21 T12 =T22 S11 S12 ¼ ð2:50Þ S21 S22 1=T22 T12 =T22
0 3
The matrix S is called the scattering matrix, or S matrix. It will play a central role in the discussion of scattering in Chapter 14. One of the many advantages of introducing the scattering matrix is that reflection and transmission coefficients can be easily expressed in terms of its elements. For example, if the particle is incident from the left, so that B00 ¼ 0, then it follows from eqn 2.49 that A00 ¼ S12 A
–10
67
Likewise, the relations between the coefficients A 0 , B 0 , A00 , and B00 in Zones II and III can be expressed as another, slightly more complex, matrix expression: ! k þ ik ðk þ ikÞe2kL eðkþikÞL 00 0 Q¼ C ¼ QC ð2:46Þ 2ik ðk þ ikÞe2ikL ðk þ ikÞe2ikL e2kL
1 0
Probability density, P (x,t )
3 2
j
–5 0 5 Displacement, x
10
Fig. 2.30 (a) The trajectory of
a wavepacket, in this case of a ‘coherent state’, a wavefunction for which the uncertainty product DpDx has its minimum value of 12 h. This wavepacket oscillates backwards and forwards with the classical frequency, and although it spreads and contracts a little with time, at the end of each period it has its initial shape and location. The numbers denote the sequence of four snapshots. (b) This wavepacket has a different composition, with the spreading more pronounced.
B ¼ S22 A
Therefore, the reflection and transmission probabilities are R ¼ jS22 j2
T ¼ jS12 j2
ð2:51Þ
Example 2.4 Properties of the S matrix
A property of the S matrix is that it is unitary (see below). Show that the unitarity of the S matrix implies that T þ R ¼ 1. Method. The unitarity of the S matrix means that
Sy S ¼ SSy ¼ 1 where Sy is the adjoint of S (the complex conjugate of its transpose): S11 S21 S11 S12 S11 S12 T S11 S21 ; Sy ¼ If S ¼ ¼ ¼ S12 S22 S21 S22 S21 S22 S12 S22 The unitarity of the S matrix is established in Further information 13. The condition T þ R ¼ 1 can be expressed in terms of the elements of the S matrix
68
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
by using eqn 2.51. We should inspect the relation and see if it is implied by the unitarity condition by writing the latter out in terms of the elements of S. Answer. In terms of the elements of the S matrix, the condition T þ R ¼ 1 is
jS12 j2 þ jS22 j2 ¼ 1 The condition SyS ¼ 1, when written out in full, is ! ! ! S11 S11 þ S21 S21 S11 S12 þ S21 S22 S11 S21 S11 S12 ¼ S12 S22 S12 S11 þ S22 S21 S12 S12 þ S22 S22 S21 S22 ! 1 0 ¼ 0 1 Comparison of the (2,2)-elements implies that S12 S12 þ S22 S22 ¼ 1 which is the same as T þ R ¼ 1. Comment. As this calculation suggests, the unitarity of the S matrix is essen-
tially a way of saying that the number of particles is preserved during the scattering event, because T þ R ¼ 1 expresses the fact that the sum of the probabilities of transmission and reflection is 1. Whenever you see ‘unitarity’ referred to, think of it as implying the conservation of probability. Conversely, if you want to ensure that probability is conserved, then you should impose the property of unitarity on the matrices you are using. Self-test 2.4. Suppose the particle flux is incident from the right of the barrier.
Define T and R in terms of the appropriate S matrix elements and confirm that T þ R ¼ 1.
PROBLEMS 2.1 Write the wavefunctions for (a) an electron travelling to the right (x > 0) after being accelerated from rest through a potential difference of (i) 1.0 V, (ii) 10 kV, (b) a particle of mass 1.0 g travelling to the right at 10 m s1. 2.2 Find expressions for the probability densities of the particles in the preceding problem. 2.3 Use the qualitative ‘wavefunction generator’ in Fig. 2.4 to sketch the wavefunctions for (a) a particle with a potential energy that decreases linearly to the right, (b) a particle with a potential energy that is constant to x ¼ 0, then falls in the shape of a semicircle to a low value to climb back to its original constant value at x ¼ L, (c) the same as in part (b), but with the dip replaced by a hump.
2.4 Express the coefficients C and D in eqn 2.6 in terms of the coefficients A and B in eqn 2.5. 2.5 Calculate the flux density (eqn 2.11) for a particle with a wavefunction with coefficients A ¼ A0 cos z and B ¼ A0 sin z, for a particle undergoing free motion in one dimension, with z a parameter, and plot Jx as a function of z. 2.6 A particle was prepared travelling to the right with all momenta between (k 12Dk) h and (k þ 12Dk) h contributing equally to the wavepacket. Find the explicit form of the wavepacket at t ¼ 0, normalize it, and estimate the range of positions, Dx, within which the particle is likely to be found. Compare the last
PROBLEMS
conclusion with a prediction based on the uncertainty principle. Hint. Use eqn 2.13 with g ¼ B, a constant, inside the range k 12Dk to k þ 12Dk and zero Relsewhere, and eqn 2.12 with t ¼ 0 for Ck. To evaluate jCkj2dt (for step) use the integral R 1 the normalization 2 1ðsin x=xÞ dx ¼ p. Take Dx to be determined by the locations where jCj2 falls to half its value at x ¼ 0. For the last part use Dpx hDk. 2.7 Sketch the form of the wavepacket constructed in Problem 2.6. Sketch its form a short time after, when t is non-zero but still small. Hint. For the second part use 2 eqn 2.13 but with eihk t/2m 1 ihk2t/2m. Use a computer to draw the wavepacket at longer times, evaluating the appropriate integrals numerically. 2.8 Repeat the evaluation that led to eqn 2.22 but do so for the case E > V. Compare your result to the transmission probability in eqn 2.23. 2.9 A particle of mass m is incident from the left on a wall of infinite thickness and which may be represented by a potential energy V. Calculate the reflection coefficient for (a) E V, (b) E > V. For electrons incident on a metal surface V ¼ 10 eV. Evaluate and plot the reflection coefficient. Hint. Proceed as in the last problem but consider only two domains, inside the barrier and outside it. The reflection coefficient is the ratio jBj2/jAj2 in the notation of eqn 2.21a. 2.10 A particle of mass m is confined to a one-dimensional box of length L. Calculate the probability of finding it in the following regions: (a) 0 x 12L, (b) 0 x 14L, (c) 1 1 2L dx x 2L þ dx. Derive expressions for a general value of n, and then specialize to n ¼ 1. 2.11 An electron is confined to a one-dimensional box of length L. What should be the length of the box in order for its zero-point energy to be equal to its rest mass energy (mec2)? Express the result in terms of the Compton wavelength, lC ¼ h/mec. 2.12 Energy is required to compress the box when a particle is inside: this suggests that the particle exerts a force on the walls. (a) On the basis that when the length of the box changes by dL the energy changes by dE ¼ FdL, find an expression for the force. (b) At what length does F ¼ 1 N when an electron is in the state n ¼ 1? 2.13 The mean position hxi of a particle in a onedimensional well can be calculated by weighting its position x by the probability that it will be found in the region dx at x, which is c2(x)dx, and then summing 1 (i.e. integrating) these values. R LShow that hxi ¼ 2L for all values of n. Hint. Evaluate 0 xc2n ðxÞdx: 2.14 The root mean square deviation of the particle from its mean position is Dx ¼ {hx2i hxi2}1/2. Evaluate this quantity for a particle in a well and show that it
j
69
approaches R L its classical value as n ! 1. Hint. Evaluate hx2i ¼ 0 x2c2(x)dx. In the classical case the distribution is uniform across the box, and so in effect c(x) ¼ 1/L1/2. 2.15 For a particle in a box, the mean value and mean square linear momentum are given by R L value of Rthe L 2 0 c pcdx and 0 c p cdx, respectively. Evaluate these quantities. Form the r.m.s. deviation Dp ¼ {h p2i h pi2}1/2 and investigate the consistency of the outcome with the uncertainty principle. Hint. Use p ¼ ( h/i)d/dx. For h p2i 2 notice that E ¼ p /2m and we already know E for each n. For the last part, form DxDp and show that DxDp 12 h, the precise form of the principle, for all n; evaluate DxDp for n ¼ 1. 2.16 Calculate the energies and wavefunctions for a particle in a one-dimensional square well in which the potential energy rises to a finite value V at each end, and is zero inside the well. Show that for any V and L there is always at least one bound level, and that as V ! 1 the solutions coincide with those in eqn 2.30. Hint. This is a difficult problem. Divide space into three zones, solve the Schro¨dinger equations, and impose the boundary conditions (finiteness and continuity of c and continuity of dc/dx across the zone boundaries: combine the latter continuity requirements into the continuity of the logarithmic derivatives ((1/c)(dc/dx) ). After some algebra arrive at ( ) k h kL þ 2 arcsin ¼ np k h ¼ ð2mEÞ1=2 ð2mVÞ1=2 Solve this expression graphically for k and hence find the energies for each value of the integer n. 2.17 (a) Confirm eqn 2.22 and eqn 2.23 for the onedimensional transition probability. (b) Demonstrate that the two expressions coincide at E ¼ V and identify the value of T at that energy. 2.18 Identify the locations of the nodes in the wavefunction with n ¼ 4 for a particle in a one-dimensional square well. 2.19 A very simple model of a polyene is the free electron molecular orbital (FEMO) model. Regard a chain of N conjugated carbon atoms, bond length RCC, as forming a box of length L ¼ (N 1)RCC. Find the allowed energies. Suppose that the electrons enter the states in pairs so that the lowest 12N states are occupied. Estimate the wavelength of the lowest energy transition. Sometimes the length of the chain is taken to be (N þ 1)RCC, allowing for electrons to spill over the ends slightly. 2.20 (a) Show that the variables in the Schro¨dinger equation for a cubic box may be separated and the overall wavefunctions expressed as X(x)Y(y)Z(z).
70
j
2 LINEAR MOTION AND THE HARMONIC OSCILLATOR
(b) Deduce the energy levels and wavefunctions. (c) Show that the functions are orthonormal. (d) What is the degeneracy of the level with E ¼ 14(h2/8mL2)? 2.21 (a) Demonstrate that accidental degeneracies can exist in a rectangular infinite square-well potential provided that the lengths of the sides are in a rational proportion. (b) What are the degeneracies when L1 ¼ lL2, with l an integer? 2.22 Find the form of the ground-state wavefunction of a particle of mass m in an infinitely deep circular square well of radius R. Hint. Separate the Schro¨dinger equation for the system; the radial wavefunctions are related to Bessel functions. 2.23 The Hermite polynomials Hv (y) satisfy the differential equation Hv00 ðyÞ 2yHv0 ðyÞ þ 2Hv ðyÞ ¼ 0 Confirm that the wavefunctions in eqn 2.41 are solutions of the harmonic oscillator Schro¨dinger equation. 2.24 Locate the nodes of the harmonic oscillator wavefunction for the state with v ¼ 6. 2.25 Confirm the expression for the normalization factor of a harmonic oscillator wavefunction, eqn 2.41. 2.26 Evaluate the matrix elements (a) hv þ 1jxjvi and hv þ 2jx2jvi of a harmonic oscillator by using the recursion relations of the Hermite polynomials. 2.27 The oscillation of the atoms around their equilibrium positions in the molecule HI can be modelled as a harmonic oscillator of mass m mH (the iodine atom is almost stationary) and force constant k ¼ 313.8 N m1. Evaluate the separation of the energy levels and predict the wavelength of the light needed to induce a transition between neighbouring levels. 2.28 What is the relative probability of finding the HI molecule with its bond length 10 per cent greater than its equilibrium value (equilibrium bond length of 161 pm) when it is in (a) the v ¼ 0 state, (b) the v ¼ 4 state? Use the information in Problem 2.27. 2.29 Calculate the values of (a) hxi, (b) hx2i, (c) h pxi, (d) h px2i for a harmonic oscillator in its ground state by evaluation of the appropriate integrals (as in Problems 2.13–2.15). Examine the value of DxDpx in the light of the uncertainty principle. Hint. Use the integrals
Z
1
2
eax dx
1
Z 0
1
2
xeax dx ¼
1 2a
Z
1
1
2
x2 eax dx ¼
1 p 1=2 2 a3
2.30 Equation 2.50 gives the form of the S matrix for a onedimensional system in which a particle is scattered from an abrupt blip in the potential energy. Write down the analogous expression for scattering from a comparable dip in the potential energy. 2.31 Show that the flux density associated with a time-dependent wavefunction C of definite energy is independent of location. Hint. Use eqn 2.10 in conjunction with the time-dependent Schro¨dinger equation to show that Jx is independent of x; that is, qJx/qx ¼ 0. 2.32 A particle of mass m is confined in a one-dimensional box of length L. The state of the particle is given by the normalized wavefunction c(x) ¼ 13c1(x) þ 13ic3(x) (79)1/ 2 c5(x) where cn(x) is a normalized particle-in-a-box wavefunction corresponding to quantum number n (eqn 2.31). (a) What will be the outcome when the energy of the particle is measured? (b) If more than one result is possible, give the probability of obtaining each result. (c) What is the expectation value of the energy? 2.33 Consider a harmonic oscillator of mass m undergoing harmonic motion in two dimensions x and y. The potential energy is given by V(x,y) ¼ 12kxx2 þ 12kyy2. (a) Write down the expression for the Hamiltonian operator for such a system. (b)What is the general expression for the allowable energy levels of the two-dimensional harmonic oscillator? (c) What is the energy of the ground state (the lowest energy state)? Hint. The hamiltonian operator can be written as a sum of operators. 2.34 Consider a particle of mass 1.00 1025 g freely moving in a (microscopic) three-dimensional cubic box of side 10.00 nm. The potential energy is zero inside the box and is infinite at the walls and outside of the box. (a) Evaluate the zero-point energy of the particle. (b) Consider the energy level that has an energy 9 times greater than the zero-point energy. What is the degeneracy of this level? Identify all the sets of quantum numbers that correspond to this energy. (The energy levels of the cubic box were deduced in Problem 2.20.)
3
Rotational motion and the hydrogen atom
Particle on a ring 3.1 The hamiltonian and the Schro¨dinger equation 3.2 The angular momentum 3.3 The shapes of the wavefunctions 3.4 The classical limit Particle on a sphere 3.5 The Schro¨dinger equation and its solution 3.6 The angular momentum of the particle 3.7 Properties of the solutions 3.8 The rigid rotor
The second class of motion we consider is rotational motion, the motion of an object around a fixed point. With this problem we encounter ‘angular momentum’, which is one of the most important topics in quantum mechanics. In this chapter we discuss rotational motion and angular momentum in terms of solutions of the Schro¨dinger equation, but we return to the topic in the next chapter and see how its properties emerge from the operators for angular momentum. This is a chapter for pictures; the next provides the algebra beneath the pictures. The material we describe here occurs throughout quantum mechanics. In particular, it crops up wherever we are interested in the motion of a particle in a central potential, in which the potential energy depends only on the distance from a single point. One example is the central potential experienced by an electron in a hydrogen atom. That problem is also exactly solvable, and we shall consider it in this chapter too.
Motion in a Coulombic field 3.9 The Schro¨dinger equation for hydrogenic atoms
Particle on a ring
3.10 The separation of the relative coordinates
As a first step, we consider the quantum mechanical description of a particle travelling on a circular ring. This problem is more general than it might seem, for as well as applying to the motion of a bead on a circle of wire, it also applies to any body rotating in a plane (for example, a compact disk, Fig. 3.1). This generality stems from the fact that any such body can be represented by a mass point moving in a circle of radius r, its radius of gyration about the centre of mass. We shall see, in fact, that the property that determines the characteristics of the rotational motion of a body is the moment of inertia, I ¼ mr2, and it is not necessary to enquire into whether the value of I for a body is that of an actual particle moving on a ring of radius r or is that of a body of mass m and radius of gyration r rotating about its own centre of mass.
3.11 The radial Schro¨dinger equation 3.12 Probabilities and the radial distribution function 3.13 Atomic orbitals 3.14 The degeneracy of hydrogenic atoms
3.1 The hamiltonian and the Schro¨dinger equation The particle of mass m travels on a circle of radius r in the xy-plane. Its potential energy is constant and taken to be zero. The hamiltonian is therefore ! h2 q2 q2 ð3:1Þ þ H¼ 2m qx2 qy2
72
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
Because the motion is confined to a circle, a simpler expression is obtained by adopting polar coordinates and writing x ¼ r cos f and y ¼ r sin f where f ranges from 0 to 2p. The laplacian r2 in two dimensions is
r
R
r = R/√2
M
Fig. 3.1 The rotational
characteristics of a uniform disk are represented by the motion of a single mass point at its radius of gyration.
Expressions for the laplacian in different coordinate systems are commonly derived in multivariable calculus books. A general expression can be found, for example, in M.L. Boas, Mathematical methods in the physical sciences, Wiley (1983).
q2 q2 q2 1 q 1 q2 þ þ ¼ þ qx2 qy2 qr2 r qr r2 qf2
ð3:2Þ
Then, with r constant so that derivatives with respect to r can be discarded, the hamiltonian is H¼
2 d2 h h2 d2 ¼ 2 2 2mr df 2I df2
ð3:3Þ
The wavefunction depends only on the angle f, so we denote it F. The Schro¨dinger equation is therefore d2 F 2IE ¼ 2 F 2 df h The general solutions are F ¼ Aeiml f þ Beiml f
ð3:4Þ ml ¼
2IE 1=2 2 h
ð3:5Þ
The quantity ml is a dimensionless number, and at this stage it is completely unrestricted in value; the significance of the subscript l will become apparent later. Example 3.1 The separation of the Schro¨dinger equation
The wavefunctions for a particle on a ring also arise in connection with a particle confined to a circular region of zero potential energy by potential walls of infinite height (a ‘circular square well’). Show that the Schro¨dinger equation is separable, and find equations for the radial and angular components. Method. We try to separate the equation by proposing a solution in the form c(r, f) ¼ R(r)F(f). The hamiltonian for the problem has only a kinetic energy contribution in the circular region where the particle may be found. It follows from the symmetry of the problem that it is sensible to express the hamiltonian in polar coordinates. The laplacian in two dimensions, which is needed to write the hamiltonian, is given in eqn 3.2. Answer. It follows from eqn 3.2 that the Schro¨dinger equation inside the well is
( ) 2 q2 c 1 qc 1 q2 c h þ þ ¼ Ec 2m qr2 r qr r2 qf2 Substitution of c ¼ RF and then division of both sides by RF gives ! h2 1 1 h2 F00 R00 þ R0 ¼E r 2m R 2mr2 F where R 0 and R00 are first and second derivatives with respect to r and F00 is the second derivative with respect to f. The 1/r2 in the second term can be
3.2 THE ANGULAR MOMENTUM
j
73
eliminated by multiplication through by r2, and after a little rearrangement the equation becomes
2mE 1 2 00 F00 r R þ rR0 þ 2 r2 ¼ F R h This equation is separable, because the left is a function only of r and the right is a function only of f. We therefore write F00 ¼ m2l F which implies that r2 R00 þ rR0 þ
2mE 2 h
r2 R ¼ m2l R
Self-test 3.1. Go on to solve the radial equation by identifying the form of the equation by reference to Chapter 9 of M. Abramowitz and I.A. Stegun, Handbook of mathematical functions, Dover (1965) or a similar source. A vector product between two vectors a and b, denoted a b, is a vector of length ab sin y, where y is the angle between the two vectors, a and b are the lengths of the two vectors, and the vector product is directed perpendicular to the plane defined by a and b. To construct the components of the vector product from the components of the two vectors a ¼ axi þ ay j þ azk and b ¼ bxi þ by j þ bzk, we expand the following determinant: i j k a b ¼ ax ay az bx by bz
¼ ay bz az by i þ ð az bx ax bz Þj
þ ax by ay bx k The expansion of a general 3 3 determinant is a b c d e f g h i ¼ ðaei þ bfg þ cdhÞ ðceg þ afh þ bdiÞ The properties of vectors are summarized in Further information 22.
Now we introduce the boundary conditions. There are no barriers to the particle’s motion so long as it remains on the ring, so there is no requirement for the wavefunctions to vanish at any point on the ring. However, because wavefunctions must be single-valued (Chapter 2), it follows that F(f þ 2p) ¼ F(f). This requirement is an example of a cyclic boundary condition. It follows that Aeiml f e2piml þ Beiml f e2piml ¼ Aeiml f þ Beiml f This relation is satisfied only if ml is an integer, for then, using Euler’s relation, e2piml ¼ 1. The boundary conditions therefore imply that ml ¼ 0, 1, 2, . . . It follows (from eqn 3.5) that the allowed energies are Eml ¼
m2l h2 2I
with ml ¼ 0, 1, 2, . . .
ð3:6Þ
3.2 The angular momentum By analogy with the discussion of wavefunctions for linear momenta p ¼ k h with opposite signs of k, it can be anticipated that opposite signs of ml correspond to opposite directions of circular motion. To confirm that this is so, we consider the z-component of the angular momentum l. The classical expression for l is i j k ð3:7Þ l ¼ r p ¼ x y z px py pz where i, j, and k are orthogonal unit vectors along the x-, y-, and z-axes, respectively. With the angular momentum written as l ¼ lxi þ lyj þ lzk, we can expand the determinant in eqn 3.7 and pick out the z-component as lz ¼ xpy ypx
ð3:8Þ
74
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM z
At this point we express the classical observable as an operator in the position representation: h q h q y lz ! x i qy i qx
ml > 0
Substitution of the polar coordinates defined above results in the expression lz ¼
q h i qf
ð3:9Þ
Now consider the effect of this operator on the wavefunction with B ¼ 0: ml < 0
lz F m l ¼ Fig. 3.2 The vector representation of angular momentum of a particle (or an effective particle) confined to a plane. Note the right-hand screw convention for the orientation of the vector.
q h Aeiml f ¼ ml hAeiml f ¼ ml hFml i qf
ð3:10Þ
This is an eigenvalue equation, and we see that the wavefunction corresponds to an angular momentum ml h. If ml > 0, then the angular momentum is positive, and if ml < 0, then the angular momentum is negative (Fig. 3.2). The remaining task is to normalize the wavefunctions. For the function with B ¼ 0, we write Z 2p Z 2p Z 2p F F df ¼ jAj2 eiml f eiml f df ¼ jAj2 df ¼ 2pjAj2 0
0
0
It follows that j A j ¼ 1/(2p)1/2, and A is conventionally chosen to be real (so the modulus bars can be dropped from this relation). It is easy to go on to show that the wavefunctions with different values of ml are mutually orthogonal (see Problem 3.4).
Wavefunction,
3.3 The shapes of the wavefunctions The physical basis of the quantization of rotation becomes clear when we inspect the shapes of the wavefunctions. The wavefunction corresponding to h is a state of definite angular momentum ml 1=2 1=2 1 1 Fm l ¼ eiml f ¼ fcos ml f þ i sin ml fg ð3:11Þ 2p 2p
0
2π
Fig. 3.3 The wavefunction must satisfy cyclic boundary conditions; only the dark curve of these three is acceptable. The horizontal coordinate corresponds to an entire circumference of the ring, and the end points should be considered to be joined.
Note that the wavefunction is complex (for ml 6¼ 0), which is another illustration of the fact that wavefunctions corresponding to definite states of motion (other than being stationary in the sense that ml ¼ 0) are complex. We shall consider explicitly only the cosine component of the function, but similar remarks apply to the sine component too: the two components are 90 out of phase. When ml is an integer, the cosine functions form a wave with an integral number of wavelengths wrapped round the circular ring. The ‘ends’ of the wave join at f and f þ 2p, and the function reproduces itself on the next circuit (Fig. 3.3). When ml is not an integer (for one of the disallowed solutions), the wavefunction has an incomplete number of wavelengths between 0 and 2p, and does not reproduce itself on the next circuit of the ring. At any point, it is double-valued, and hence must be rejected.
3.3 THE SHAPES OF THE WAVEFUNCTIONS Wavefunction, Φ
Fig. 3.4 One wavefunction for a particle on a ring (with ml ¼ 1). Only the real part is shown.
Wavefunction, Φ()
Im Φ
0
2π
Re Φ
Fig. 3.5 A wavefunction corresponding to a definite state of motion is complex. The real and imaginary components shown here correspond to ml ¼ þ1. Note that the real component seems to chase the imaginary one. The state with ml ¼ 1 has the imaginary component shifted in phase by p (that is, the component is multiplied by 1).
j
75
A glance at the expression for the energy shows that all the levels except the lowest (ml ¼ 0) are doubly degenerate: because Eml / m2l , the states þ j ml j and j ml j have the same energy. This degeneracy stems from the fact that the particle can travel in either direction around the ring with the same magnitude of angular momentum, and hence with the same kinetic energy. The ground state is non-degenerate because when ml ¼ 0 the particle is stationary and the question of alternative directions of travel does not arise. There are several ways of depicting the wavefunctions. The simplest procedure is to plot the real part of F on the perimeter of the ring (Fig. 3.4). It should be noted that in general the wavefunction is complex, and so it has real and imaginary components displaced by 90 . It is therefore easier to unwrap the ring into a straight line in the range 0 f 2p and to plot the wavefunctions on this line (Fig. 3.5). Drawing the two components helps to remind us that although the amplitude varies from point to point, the probability density is uniform: 1=2 1=2 1 1 1 ð3:12Þ eiml f eiml f ¼ jFml j2 ¼ 2p 2p 2p In a state of definite angular momentum, the particle is distributed uniformly round the ring: certainty in the value of the angular momentum implies total uncertainty in the location of the particle. A second point is that as the energy and the angular momentum increase, so the number of nodes in the real and imaginary components of the wavefunction increases too. This is an example of the behaviour we have already discussed: as the number of nodes is increased, the wavefunction is buckled backwards and forwards more sharply to fit into the perimeter of the ring, and consequently the kinetic energy of the particle increases. A further point that will prove to be of significance later is that the wavefunctions have the following symmetry properties: 1=2 1=2 1 1 eiml ðfþpÞ ¼ eiml f ðeip Þml ¼ ð1Þml Fml ðfÞ Fml ðf þ pÞ ¼ 2p 2p ð3:13Þ That is, points separated by 180 across the diameter of the ring have the same amplitude but differ in sign if ml is odd. A particle on a ring has no zero-point energy (E0 ¼ 0). The particle can satisfy the cyclic boundary conditions without its wavefunction needing to be curved (when ml ¼ 0, F is a constant), so one possible state has zero kinetic energy. The same argument is sometimes expressed in terms of the uncertainty principle in the form that as the particle may be anywhere in an infinite range of angles, its angular momentum can be specified precisely, and may be zero. However, great care must be taken when applying the uncertainty principle to periodic variables. In such cases it is appropriate to use more elaborate forms of the observables than simply f itself, and then1 Dlz D sin f 12 hjhcos fij
ð3:14Þ
.......................................................................................................
1. See P. Carruthers and M.M. Nieto in Rev. Mod. Phys. 411, 40 (1968).
76
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
3.4 The classical limit When a particle is prepared with an energy that is imprecise in comparison with the energy-level separations, as when a macroscopic disk is set spinning, the correct description of the system is as a superposition of angular momentum (and energy) eigenfunctions (eqn 3.11). The superposition results in a wavepacket. The amplitude of the wavepacket may represent the location of the actual particle or of a point representing the mass of the spinning disk. Because each component has the form (recall eqn 1.31) 1=2 1 2 eiml fiml ht=2I ð3:15Þ Cml ðf; tÞ ¼ 2p
Fig. 3.6 A wavepacket formed from the superposition of many angular momentum eigenfunctions moves round the ring like the location of a classical particle. However, it also spreads with time.
. the point of maximum interference rotates around the ring (Fig. 3.6) This motion corresponds to the classical description of a rotating body. Rotating motion in classical physics is normally denoted by a vector that represents the state of angular momentum of the body. For motion confined to the xy-plane, the vector lies parallel to the z-axis (eqn 3.7). The length of the vector represents the magnitude of the angular momentum, and its direction indicates the direction of motion. The right-hand screw convention is adopted: a vector pointing towards positive z represents clockwise rotation seen from below (as in Fig. 3.2). A vector pointing towards negative z represents motion in a counter-clockwise sense seen from below. The same representation can be used in quantum mechanics, the only difference being that in this case the length of the vector is confined to discrete values corresponding to the allowed values of ml whereas in classical physics the length is continuously variable.
Particle on a sphere Now we consider the case of a particle free to move over the surface of a sphere. The mass point can be an actual particle or a point in a solid body that represents the motion of the whole body. For example, a solid uniform sphere of mass m and radius R can be represented by the motion of a single point of mass m at a distance r ¼ (25)1/2R (the radius of gyration) from the centre of the sphere. This problem will build on the material covered in the previous section and prove to be the foundation for many applications in later chapters.
3.5 The Schro¨dinger equation and its solution The potential energy of the particle is a constant taken to be zero, so the hamiltonian for the problem is simply H¼
2 2 h r 2m
ð3:16Þ
3.5 THE SCHRO¨ DINGER EQUATION AND ITS SOLUTION z
j
77
It is convenient to mirror the spherical symmetry of the problem by expressing the derivatives in terms of spherical polar coordinates (Fig. 3.7):
x ¼ r sin y cos f
r
y
y ¼ r sin y sin f
z ¼ r cos y
Standard manipulation of the differentials leads to the following expression for the laplacian operator: 1 q2 1 r þ 2 L2 2 r qr r Two equivalent, alternative forms are
r2 ¼
x Fig. 3.7 Spherical polar coordinates. The angle y is called the colatitude and the angle f is the azimuth.
ð3:17Þ
ð3:18aÞ
r2 ¼
1 q 2 q 1 r þ L2 r2 qr qr r2
ð3:18bÞ
r2 ¼
q2 2 q 1 þ L2 þ qr2 r qr r2
ð3:18cÞ
The legendrian, L2, the angular part of the laplacian, is defined as L2 ¼
1 q2 1 q q sin y þ 2 2 sin y qy qy sin y qf
ð3:19Þ
The condition that the particle is confined to the surface of fixed radius is equivalent to ignoring the radial derivatives, so we retain only the legendrian part of the laplacian and treat r as a constant. The hamiltonian is therefore 2 h L2 ð3:20Þ 2mr2 Then, because the moment of inertia is I ¼ mr2, the Schro¨dinger equation we have to solve is 2IE c ð3:21Þ L2 c ¼ h2 H¼
z
x
y
Fig. 3.8 The motion of a particle on
the surface of a sphere is like its motion on a stack of rings with the ability to pass between the rings.
where c is a function of the angles y and f. There are three ways of solving this second-order partial differential equation. One is to realize that the functions should resemble the solutions we have already found for the particle on a ring, for from one point of view (from any point of view, in fact) a sphere can be regarded as a stack of rings (Fig. 3.8). The difference is that for a sphere, the particle can travel from ring to ring. This view suggests that the wavefunction ought to be separable and of the form c(y,f) ¼ Y(y)F(f). Indeed, it is easy to verify that the Schro¨dinger equation does separate, and that the component equation for F is d2 F ¼ constant F df2 This equation is the same as the one for a particle on a ring, and the cyclic boundary conditions are the same. The solutions are therefore the same as before, and are specified by the quantum number ml, with integral values. The equation for Y is much more involved and its solution by elementary techniques is cumbersome (it is given in Further information 9). The second method of solution is to avoid dealing with the Schro¨dinger equation
78
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
Table 3.1 Spherical harmonics l
ml
Ylml(, )
0
0
1/21/2
1
0
1/2 1 cos 2(3/p) 1/2
1 2
0 1 2
3
0 1 2 3
(3/2)
y
sin e i
1/2 1 (3 cos2 y 1) 4(5/p) 12(15/2p)1/2 cos y sin y e if 1/2 1 sin2 y e 2if 4(15/2p) 1/2 1 (2 5 sin2 y) cos y 4(7/p) 1/2 1 8(21/p) (5 cos2y 1) sin y e if 1/2 1 cos y sin2 y e 2if 4(105/2p) 1/2 1 8(35/p) sin3 y e 3if
directly, and to use the properties of the angular momentum operators themselves. The latter is a succinct and powerful approach, and will be described in Chapter 4. The third method of solution is to make the straighforward claim that we recognize eqn 3.21 as a well-known equation in mathematics, so that we can simply refer to tables for its solutions.2 Indeed, solution by recognition is in fact the way that many differential equations are tackled by professional theoreticians, and it is a method not to be scorned! As we show in Further information 9, the solutions of eqn 3.21 are the functions called spherical harmonics, Ylml (y, f). These highly important funcons satisfy the equation L2 Ylml ¼ lðl þ 1ÞYlml
ð3:22Þ
where the labels l and ml have the following values: l ¼ 0, 1, 2, . . .
ml ¼ l, l 1, . . . , l
Equation 3.22 has the same form as eqn 3.21, so the wavefunctions c are proportional to the spherical harmonics. The spherical harmonics are composed of two factors: Ylml ðy, fÞ ¼ Ylml ðyÞFml ðfÞ
ð3:23Þ
in accord with the separability of the Schro¨dinger equation. The functions F are the same as those already described for a particle on a ring. The functions Y are called associated Legendre functions. Table 3.1 lists the first few spherical harmonics.
.......................................................................................................
2. This is in practice a common way of solving differential equations, and the Handbook of mathematical functions mentioned in Example 3.1 is an excellent source of the appropriate information. It is an ideal desert-island book for shipwrecked quantum chemists.
3.6 THE ANGULAR MOMENTUM OF THE PARTICLE
j
79
Example 3.2 How to confirm that a spherical harmonic is a solution
Confirm that the spherical harmonic Y10 is a solution of eqn 3.22. Method. The direct method is to substitute the explicit expression for the spherical harmonic, taken from Table 3.1, into the left-hand side of eqn 3.22 and to verify that it is equal to the expression given on the right-hand side. The expression for the legendrian operator is given in eqn 3.19; because Y10 is independent of f (see Table 3.1), the partial derivatives with respect to f are zero, and we need consider only the derivatives with respect to y. Answer. It follows from Table 3.1 (writing N for the normalization constant)
that 1 q q 1 d sin y N cos y ¼ N sin2 y sin y qy qy sin y dy 1 sin y cos y ¼ 2Y10 ¼ 2N sin y
L2 Y10 ¼
This result is consistent with eqn 3.22 when l ¼ 1. Self-test 3.2. Confirm that Y21 is a solution.
Comparison of eqns 3.21 and 3.22 shows that the energies of the particle are confined to the values 2 h ð3:24Þ 2I The quantum number l is a label for the energy of the particle. Notice that Elml is independent of the value of ml. Therefore, because for a given value of l there are 2l þ 1 values of ml, we conclude that each energy level is (2l þ 1)fold degenerate. Elml ¼ lðl þ 1Þ
3.6 The angular momentum of the particle The quantum numbers l and ml have a further significance. The rotational energy of a spherical body of moment of inertia I and angular velocity o is given by classical physics as E ¼ 12Io2. Because the magnitude of the angular momentum is related to the angular velocity by l ¼ Io, this energy can be expressed as E ¼ l2/2I. Comparison of this expression with the one in eqn 3.24 shows that Magnitude of the angular momentum ¼ flðl þ 1Þg1=2 h
ð3:25Þ
Thus, the magnitude of the angular momentum is quantized in quantum mechanics. Indeed, l is called the angular momentum quantum number. This result will be confirmed formally in Chapter 4. The spherical harmonics are also eigenfunctions of lz: h q eiml f Ylml pffiffiffiffiffiffi ¼ ml hYlml ð3:26Þ lz Ylml ¼ i qf 2p This result too will be derived more formally in Chapter 4. We see from it that ml specifies the component of the angular momentum around the z-axis,
80
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM z
the contribution to the total angular momentum that can be ascribed to rotation around that axis. However, because ml is restricted to certain values, the z-component of the angular momentum is also restricted to 2l þ 1 discrete values for a given value of l. This restriction of the component of angular momentum is called space quantization. The name stems from the vector representation of angular momentum in which the angular momentum is represented by a vector of length {l(l þ 1)}1/2 orientated so that its component h for the angular momentum are on the z-axis is of length ml; units of implicitly understood. The angle y from the z-axis is given by geometry
+2 √6
ml +1
√6 √6
0
cos y ¼
√6
–1 √6
–2
Fig. 3.9 The five (that is, 2l þ 1) allowed orientations of the angular momentum with l ¼ 2. The length of the vector is {l(l þ 1)}1/2, which in this case is 61/2.
z
ml +2 +1
0
–1
–2
Fig. 3.10 To represent the fact
that if the z-component of angular momentum is specified, the x- and y-components cannot in general be specified, the angular momentum vector is supposed to lie at an indeterminate position on one of the cones shown here (for l ¼ 2).
ml flðl þ 1Þg1=2
ð3:27Þ
The vector can adopt only 2l þ 1 orientations (Fig. 3.9), in contrast to the classical description in which the orientation of the rotating body is continuously variable. The quantum numbers l and ml do not enable us to specify the x- and y-components of the angular momentum. Indeed, as we shall see later (Section 4.1), because the operators corresponding to these components do not commute with the operator for the z-component, these components cannot in general be specified if the z-component is known. Therefore, a better representation of the states of angular momentum of a body is in terms of the cones shown in Fig. 3.10, in which no attempt is made to display any components other than the z-component. At this stage you should not think of the angular momentum vector as sweeping around the cones but simply as lying at some unspecified position on them. It is a feature of space quantization that the angular momentum vector cannot lie exactly parallel to an arbitrarily specified z-axis; if it could, then we would be able to specify (as zero) the x- and y- components. Its maximum h. z-component is lh, which in general is less than its magnitude, {l(l þ 1)}1/2 Only for very large values of l (in the classical limit) is {l(l þ 1)}1/2 l, and then rotation can take place around a single axis. Example 3.3 The quantization of angular momentum for a macroscopic body
A solid ball of mass 250 g and radius 4.0 cm is spinning at 5.0 revolutions per second. Estimate the value of l and the minimum angle its angular momentum vector can make with respect to a selected axis. Method. We need to calculate first its angular momentum, Io, and use the
expression I ¼ mr2, with r the radius of gyration given in the text for a solid sphere of radius R, which is r ¼ (25)1/2R. Then identify l by setting the calculated value of angular momentum equal to {l(l þ 1)}1/2 h. The minimum angle can be obtained by trigonometry using eqn 3.27 for a general value of ml and then setting ml ¼ l. Answer. The angular velocity of the ball is o ¼ 2pn with n ¼ 5.0 s1. Its
moment of inertia is I ¼ (25)mR2, so its angular momentum is (45)pnmR2. We set {l(l þ 1)}1/2h equal to this quantity: flðl þ 1Þg1=2 h ¼ ð45ÞpnmR2
3.7 PROPERTIES OF THE SOLUTIONS
j
81
Because l 1, it follows that l(l þ 1) l2, and therefore that l
4pnmR2 5h
Insertion of the numerical values gives l 4.7 1031. Using eqn 3.27, for ml ¼ l and l 1 we can write cos y ¼
l flðl þ 1Þg
1=2
¼
1 ð1 þ 1=lÞ
1=2
¼
1 1 ¼ 1 þ 1 þ 1=2l þ 2l
where we have used Taylor series expansions for (1 þ x)1/2 and 1/(1 þ x). Because y 1, we can equate this expression with the Taylor series expansion cos y ¼ 1 12y2 þ . It follows that y 1=l1=2 ¼ 1:5 1016 rad Comment. This angle is virtually zero. Hence a macroscopic object can rotate Positive amplitude
effectively solely around a single specified axis. Self-test 3.3. Show that the difference between the angles y for the vectors with ml ¼ l and ml ¼ l 1 becomes zero as l becomes infinite.
Nodal plane
3.7 Properties of the solutions Negative amplitude Fig. 3.11 One representation of the
wavefunction of a particle on a sphere (with l ¼ 1, ml ¼ 0) plots the function in terms of a height above or below the surface of the sphere.
z
+
-
Fig. 3.12 In another representation of
the same wavefunction as in the preceding illustration, the function is plotted along a radius to the point in question. In this case, the resulting surface consists of two touching spheres.
The wavefunctions for a particle on a sphere—the spherical harmonics—can be represented diagramatically in a variety of ways. The most cumbersome method is to plot the amplitude of the function relative to the surface of the sphere, by analogy with the wavefunctions for a particle on a ring (Fig. 3.11). It is more convenient, however, to plot the amplitudes of the spherical harmonics as a surface, the distance from the origin indicating the amplitude at that orientation (Fig. 3.12). The spherical harmonics are complex functions for ml 6¼ 0, and the diagrams show only their real components. As for the particle on a ring, the complex function consists of a real and an imaginary component, the latter being the same shape as the former but rotated by 90 around the z-axis. An example is shown in Fig. 3.13. This illustration is included to emphasize the point that if ml is specified, then the azimuthal distribution of the particle (the distribution with respect to the azimuth f) is uniform: it is impossible to specify the azimuthal location of a particle with a well-defined component of angular momentum around the z-axis. Figure 3.14 shows the probability densities j Ylml j2 for l ¼ 0, 1, and 2 and the azimuthal uniformity is clearly apparent. Notice too how the distribution shifts towards the equator as j ml j approaches l. This change corresponds to a reduced tilt in the plane of classical rotation. For each spherical harmonic Ylml, there are l angular nodes or distinct angles (to modulo p) for which the probability density vanishes. This is also evident from Table 3.1. For example, Y10 has a nodal xy-plane (y ¼ 12p) whereas Y1 1 has a node along the z-axis (y ¼ 0). (For the former, y ¼ 32p is not considered a second angular node just as y ¼ p is not considered a second angular node for the latter.)
82
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
Imaginary component
z
y +
–
+
x l = 0, ml = 0
– Real component Fig. 3.13 The wavefunctions
corresponding to l ¼ 1, ml ¼ 1 are complex, with real and imaginary components like those shown here. The direction of motion is determined by the relative phases of the two components: the real chases the imaginary.
l = 1, ml = 0
l = 2, ml = 0
l = 1, ml = ±1
l = 2, ml = ±1
l = 2, ml = ±2
Fig. 3.14 The boundary surfaces for j c j 2 corresponding to l ¼ 0, 1, 2 and the allowed
values of j ml j in each case. z
x
y
It should be noticed that there is no zero-point energy (E00 ¼ 0) because the wavefunction need not be curved (relative to the surface of the sphere); indeed, Y00 is a constant and all its derivatives are zero. The classical description of a rotating particle is achieved when the particle is set rotating with an imprecisely defined energy. In that case, its wavefunction is a wavepacket formed from a superposition of the spherical harmonics. This wavepacket moves in accord with the predictions of classical physics and migrates through all angles, but spreads with time (Fig. 3.15).
Fig. 3.15 The motion of a
wavepacket on the surface of a sphere. As the wavepacket traces out the path like that of a classical particle, it also spreads.
3.8 The rigid rotor It is convenient at this point to introduce a variation on the topic of a particle on a sphere, to see how the same results apply to a body made up of two masses m1 and m2 at a fixed separation R. We have seen that any rigid object will be described by the same equations as for a single effective particle, but it is appropriate to present the argument more formally. As we shall see, the separation of variables technique is the key.
3.8 THE RIGID ROTOR
j
83
The hamiltonian for two particles moving in free space is H¼
2 2 h h2 2 r1 r 2m1 2m2 2
ð3:28Þ
where r2i differentiates with respect to the coordinates of particle i. As we show in Further information 4, this expression may be transformed by using 1 2 1 2 1 1 r1 þ r2 ¼ r2cm þ r2 m1 m2 m m where m ¼ m1 þ m2 and 1 1 1 ¼ þ m m1 m2
ð3:29Þ
The quantity m is called the reduced mass of the system; the subscript ‘cm’ on the first laplacian on the right indicates that the derivatives are with respect to the centre of mass coordinates of the joint system, and the absence of subscripts on the second laplacian indicates that it is composed of derivatives with respect to the relative coordinates of the pair. At this stage, the Schro¨dinger equation has become
2 2 h h2 rcm C r2 C ¼ Etotal C 2m 2m
ð3:30Þ
This equation can be separated into equations for the motion of the centre of mass and for the relative motion of the particles. To do so we write C ¼ ccmc, and by the same arguments as we have used several times before, find that the two factors separately satisfy the equations
2 2 h r c ¼ Ecm ccm 2m cm cm
ð3:31aÞ
2 2 h r c ¼ Ec 2m
ð3:31bÞ
with Etotal ¼ Ecm þ E. The first of these equations should be recognized as the translational motion of a free particle of mass m, which we solved in Chapter 2, with coordinates given by the centre of mass of the particle. The second equation needs a little more work, for although it looks as simple as the first equation, the fact that R is a constant must be taken into account by working in spherical polar coordinates. Because the separation R of the two particles is constant (for a rigid rotor), the derivative with respect to the radial coordinate plays no role in eqn 3.18. Consequently, only the legendrian component need be retained, and we obtain
2 h L2 c ¼ Ec 2mR2
ð3:32Þ
At this stage we write I ¼ mR2
ð3:33Þ
and obtain exactly the equation we have already considered (eqn 3.21). The solutions of this equation require two quantum numbers playing
84
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
the role of l and ml, and for the rigid rotor it is common to use J and MJ. The wavefunctions of the diatomic rigid rotor are the spherical harmonics YJMJ, and the energy levels are EJMJ ¼ Jð J þ 1Þ
h 2 2I
ð3:34Þ
with J ¼ 0, 1, 2, . . . and MJ ¼ 0, 1, . . . , J. Note that because the energy is independent of MJ and there are 2J þ 1 values of MJ for each value of J, each energy level is (2J þ 1)-fold degenerate. All the other features of the particle on a sphere apply equally to the rigid diatomic rotor, including the quantization of the angular momentum and space quantization.
Motion in a Coulombic field z
x
y
Fig. 3.16 The motion of a particle
in a central field of force is like its motion on a stack of spheres with the ability to pass between the spheres.
The motion of an electron in a Coulombic field, one in which the potential varies as 1/r, is of central importance in chemistry because it includes the structure of hydrogenic atoms, or one-electron species with arbitrary atomic number Z (Z ¼ 1 for hydrogen itself). Most of the work of solving the Schro¨dinger equation has in fact already been done, for the motion can be regarded as that of an electron on a series of concentric spheres (Fig. 3.16). It follows that the wavefunctions can be expected to contain factors that correspond to the motion of a particle on a sphere. The additional work we must do is to account for the radial dependence of the motion, the extra degree of freedom that allows the electron to travel between the nested spherical surfaces.
3.9 The Schro¨dinger equation for hydrogenic atoms The hamiltonian for the two-particle electron–nucleus system is H¼
h 2 2 h2 Ze2 re r2N 2me 2mN 4pe0 r
ð3:35Þ
where me is the mass of the electron, mN is the mass of the nucleus, and r2e and r2N are the laplacian operators that act on the electron and nuclear coordinates, respectively. The quantity e0 is the vacuum permittivity. Apart from the Coulombic potential energy term, this hamiltonian is the same as we considered for the two-particle rotor. When we convert to centre-of-mass and relative coordinates, the potential energy term remains unchanged because it depends only on the separation of the particles. Therefore, we can use the work in Further information 4 to write H¼
2 2 h h2 Ze2 rcm r2 2m 2m 4pe0 r
ð3:36Þ
where m ¼ me þ mN and the reduced mass is given by eqn 3.29. The resulting Schro¨dinger equation is separable on account of the dependence of the potential energy on the particle separation alone, and by the same argument
3.11 THE RADIAL SCHRO¨ DINGER EQUATION
j
85
as above, the Schro¨dinger equation for the relative motion of the electron and nucleus is
2 2 h Ze2 r c c ¼ Ec 2m 4pe0 r
ð3:37Þ
The other component of the Schro¨dinger equation is that for the translational motion of the atom as a whole, and we do not need to consider it further. Unlike the rigid rotor, the electron and nucleus are not constrained to have a fixed separation. We have to include the radial derivative in the laplacian, and so write the Schro¨dinger equation as 1 q2 1 2 Ze2 m 2mE c ð3:38Þ c¼ rc þ 2 L c þ r qr2 r 2pe0 h2 h2 r
3.10 The separation of the relative coordinates We have anticipated that the Schro¨dinger equation for the relative motion will be separable into angular and radial components, with the former being the equation for a particle on a sphere. We therefore attempt a solution of the form cðr, y, fÞ ¼ RðrÞYðy, fÞ
ð3:39Þ
where Y is a spherical harmonic. When this trial solution is substituted into the Schro¨dinger equation and we use L2Y ¼ l(l þ 1)Y, it turns into 1 q2 lðl þ 1Þ Ze2 m 2mE RY ¼ RY rRY RY þ r qr2 r2 2pe0 h2 h2 r The function Y may be cancelled throughout, and that leaves an equation for the radial wavefunction, R: 1 d2 ðrRÞ Ze2 m lðl þ 1Þ 2mE R ¼ R þ r dr2 r2 2pe0 h2 h2 r At this stage we write u ¼ rR, and so obtain ) ( d2 u 2m Ze2 lðl þ 1Þ h2 2mE u u ¼ þ dr2 4pe0 r 2mr2 h2 h2
ð3:40Þ
This is the one-dimensional Schro¨dinger equation in the coordinate r that would have been obtained if, instead of the Coulomb potential, we had used an effective potential energy Veff: Veff ¼
Ze2 lðl þ 1Þ h2 þ 2 4pe0 r 2mr
ð3:41Þ
3.11 The radial Schro¨dinger equation The effective potential energy may be given a simple physical interpretation. The first part is the attractive Coulomb potential energy. The second part is a repulsive contribution that corresponds to the existence of a centrifugal force that impels the electron away from the nucleus by virtue
86
Effective potential energy, Veff
0
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
l≠0
l=0
Radius, r Fig. 3.17 The effective potential
experienced by an electron in a hydrogen atom. When l > 0 there is a centrifugal contribution to the potential that prevents the close approach of the electron to the nucleus, as it increases more rapidly (as 1/r2) than the Coulomb attraction (which varies as 1/r).
of its motion. When l ¼ 0 the electron has no orbital angular momentum and the force—now solely the Coulombic force—is everywhere attractive. The potential energy for this special case is everywhere negative (Fig. 3.17). When l > 0 the electron possesses an orbital angular momentum which tends to fling it away from the vicinity of the nucleus, and there is a competition between this effect and the attractive part of the potential. At very short distances from the nucleus, the repulsive component tends more strongly to infinity (as 1/r2) than the attractive part (which varies as 1/r), and the former dominates. The two effective potentials (for l ¼ 0 and l 6¼ 0) are qualitatively quite different near r ¼ 0, and we shall investigate them separately. When l ¼ 0, the repulsive part of the effective potential energy is absent and the potential is everywhere attractive, even close to r ¼ 0. When r is close to zero, the magnitude of the potential energy is locally so much larger than E that the latter may be neglected in eqn 3.40. The equation then becomes d2 u 2m Ze2 u 0 for l ¼ 0 and r 0 þ dr2 4pe0 r h2 A solution of this equation is u Ar þ Br2 þ higher-order terms as can be verified by substitution of the solution and taking the limit r ! 0.3 Therefore, close to r ¼ 0 the radial wavefunction itself has the form R ¼ u/r A, which may be non-zero; that is, when l ¼ 0, there may be a non-zero probability of finding the electron at the nucleus. When l 6¼ 0, the large repulsive component of the effective potential energy of the electron at distances close to the nucleus has the effect of excluding it from that region. In classical terms, the centrifugal force on an electron with non-zero angular momentum is strong enough at short distances to overcome the attractive Coulomb force. When l 6¼ 0 and r is close to zero, eqn 3.40 becomes d2 u lðl þ 1Þ u0 dr2 r2
ð3:42Þ
because 1/r2 is the dominant term. The solution has the form B for l 6¼ 0 and r 0 rl as can be verified by substitution. Because u ¼ rR, at r ¼ 0 we know that u ¼ 0; so it follows that B ¼ 0. Therefore, the radial wavefunction has the form u R ¼ Arl for l 6¼ 0 and r 0 r This function implies that the amplitude is zero at r ¼ 0 for all wavefunctions with l 6¼ 0, and that the electron described by such a wavefunction will not be found at the nucleus. The radial wavefunction does not have a node at r ¼ 0 as a node is defined as a point where a function passes through zero. u Ar lþ1 þ
.......................................................................................................
3. The coefficients A and B are related by B ¼ AmZe2/4pe0 h2.
3.11 THE RADIAL SCHRO¨ DINGER EQUATION
j
87
Example 3.4 The asymptotic form of atomic wavefunctions at large distances
Show that at large distances from the nucleus, bound-state atomic wavefunctions decay exponentially towards zero. Method. We need to identify the terms in eqn 3.40 that survive as r ! 1, and then solve the resulting equation. When solving such asymptotic equations, the solutions should also be tested in the limit r ! 1. Answer. When r ! 1, eqn 3.40 reduces to
d2 u 2mE u ’ dr2 h2 (The sign ’ means ‘asymptotically equal to’.) However, because u ¼ rR, in the same limit this equation becomes d2 u d2 d2 R dR d2 R ’ r ¼ rR ¼ r þ 2 dr2 dr2 dr2 dr dr2 Hence d2 R 2mE 2mjEj ’ R ¼ þ R dr2 h2 h2 where we have made use of the fact that E < 0 for bound states. This equation is satisfied (asymptotically) by 2 R ’ eð2mjEj=h Þ
1=2
r
The alternative solution, with a positive exponent, is not square-integrable and so can be rejected. Hence, we can conclude that the wavefunction decays exponentially at large distances. Comment. All atomic wavefunctions, even those for many-electron atoms,
decay exponentially at large distances. Self-test 3.4. Show that the unbound states (for which E > 0) are travelling waves at large distances from the nucleus. 2 ½R ’ e ið2mjEj=h Þ
1=2
r
Explicit solutions of the radial wave equation can be found in a variety of ways. The most elementary method of solution is given in Further information 8. As explained there, the acceptable solutions are the associated Laguerre functions; the solutions are acceptable in the sense of being wellbehaved and corresponding to states of negative energy (bound states of the atom). The first few hydrogenic wavefunctions are listed in Table 3.2.4 They consist of a decaying exponential function multiplied by a simple polynomial in r. Each one is specified by the labels n and l, with n ¼ 1, 2, . . .
l ¼ 0, 1, . . . , n 1
.......................................................................................................
4. See M. Abramowitz and I.A. Stegun, Handbook of mathematical functions, Dover (1965), Chapter 22.
88
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
Table 3.2 Hydrogenic radial wavefunctions n
l
Orbital
Rnl(r)
1
0
1s
(Z/a)3/22e /2
2
0
2s
(Z/a)3/2(1/8)1/2(2 )e /2
1
2p
(Z/a)3/2(1/24)1/2 e /2
0
3s
(Z/a)3/2(1/243)1/2(6 6 þ 2)e /2
1
3p
(Z/a)3/2(1/486)1/2(4 ) e /2
2
3d
(Z/a)3/2(1/2430)1/2 2e /2
3
¼ (2Z/na)r with a ¼ 4"0h2/ e2. For an infinitely heavy nucleus, ¼ me and a ¼ a0, the Bohr radius. Relation to associated Laguerre functions: n o 3 ðnl1Þ! r=2 Rnl ðrÞ ¼ 2Z rl L2lþ1 na nþl ðrÞe 2n½ðnþlÞ! 3
Some of the radial wavefunctions are plotted as functions of r ¼ 2Zr/na0 in Fig. 3.18, where a0 is the Bohr radius:5 a0 ¼
4pe0 h2 me e2
ð3:43Þ
The numerical value of a0 is approximately 52.9 pm (see inside front cover). Note that the functions with l ¼ 0 are non-zero (and finite) at r ¼ 0, whereas all the functions with l 6¼ 0 are zero at r ¼ 0. Each radial wavefunction has n l 1 nodes (the zero amplitude at r ¼ 0 for functions with l 6¼ 0 are not nodes; recall the definition in Section 2.12). The locations of these nodes are found by determining where the polynomial in the associated Laguerre function is equal to zero. Illustration 3.1 Locating nodes
The zeros of the function with n ¼ 3 and l ¼ 0 occur where 2Z r 6 6r þ r2 ¼ 0 with r ¼ 3a0 p The zeros of this polynomial occur at r ¼ 3 3, which corresponds to p r ¼ (3 3)(3a0/2Z).
Insertion of the radial wavefunctions into eqn 3.40 gives the following expression for the energy: ! Z2 me4 1 n ¼ 1, 2; . . . ð3:44Þ En ¼ 2 n2 2 2 32p e0 h .......................................................................................................
5. In a precise calculation, the Bohr radius a0, which depends on the mass of the electron, should be replaced by a, in which the reduced mass m appears instead. Very little error is introduced by using a0 in place of a in this and the other equations in this chapter.
3.11 THE RADIAL SCHRO¨ DINGER EQUATION 2
(c)
0.3
0.05
1s
R /(Z /a0)3/2
R /(Z /a0)3/2
1.5
1.0
0.2
0.1 0
0 0
2
4
–0.1
6
0.8
–0.05 0
5
10
15
(d) 0.15
0
5
10
15
10
15
(f) 0.05 0.04
0.6
3d R /(Z /a0)3/2
0.1
0.4
R /(Z /a0)3/2
R /(Z /a0)3/2
3p
0
3s
0.5
(b)
89
(e) 0.1
0.4
R /(Z /a0)3/2
(a)
j
2s 0.2
0.03 0.02
0.05 2p
0 –0.2 0
5
10
15
0
0
5
10
0.01
15
0
0
5
Fig. 3.18 Hydrogenic radial wavefunctions: (a) 1s, (b) 2s, (c) 3s, (d) 2p, (e) 3p, (f) 3d.
The same values are obtained whatever the value of l or ml. Therefore, in hydrogenic atoms (but not in any other kind of atom) the energy depends only on the principal quantum number n and is independent of the values of l and ml; therefore, each level, as discussed below in Section 3.14, is n2-fold degenerate (that being the total number of wavefunctions for a given n). This degeneracy is peculiar to the Coulomb potential in a non-relativistic system, and we shall return to it shortly. The roles of the quantum numbers in the hydrogen atom should now be clear, but may be summarized as follows: 1. The principal quantum number, n, specifies the energy through eqn 3.44 and controls the range of values of l ¼ 0, 1, . . . , n 1; it also gives the total number of orbitals with the specified value of n as n2 and gives the total number of radial and angular nodes as n 1. 2. The orbital angular momentum quantum number, l, specifies the orbital angular momentum of the electron through eqn 3.25, and determines the number of orbitals with a given n and l as 2l þ 1. There are l angular nodes in the wavefunction; the number of radial nodes is n l 1. 3. The magnetic quantum number, ml, specifies the component of orbital h (see eqn 3.26) and, for a angular momentum of an electron through ml given n and l, specifies an individual one-electron wavefunction.
j
90
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM z
3.12 Probabilities and the radial distribution function r
x
dr
The complete wavefunctions of the electron in a hydrogenic atom have the form cnlml ¼ Rnl Ylml
y
Fig. 3.19 The radial distribution
function gives the probability that an electron will be found anywhere between two concentric spheres with radii that differ by dr.
0.6
where the Rnl are related to the (real) associated Laguerre functions and the Ylml are the (in general, complex) spherical harmonics. The probability of finding an electron in a volume element dt ¼ r2siny dydfdr at a point specified by the spherical polar coordinates (r,y,f) when the state of the electron is described by the wavefunction cnlml is j cnlml(r,y,f) j 2 dt. Although the wavefunction gives the probability of finding an electron at a specified location, it is sometimes more helpful to know the probability of finding the particle at a given radius regardless of the direction. This probability is obtained by integration over the volume contained between two concentric spheres of radii r and r þ dr (Fig. 3.19): Z p Z 2p Z c 2 dt ¼ R2nl jYlml j2 r2 sin y drdydf ð3:45Þ PðrÞ dr ¼ nlml surface
0.4
0
0
P /(Z /a0)3
The spherical harmonics are normalized to 1 in the sense that Z p Z 2p jYlml j2 sin y dydf ¼ 1 0
0.2
0
Therefore, PðrÞdr ¼ R2 r2 dr
0
0
2 r /a0
4
Fig. 3.20 The radial distribution
function for a 1s-electron. The function passes through a maximum at the Bohr radius, a0.
Equation 3.47 represents the most probable radius. Recall from the calculus that maxima, minima, and inflection points of a function correspond to points of vanishing first derivative. Evaluation of the second derivative of the function allows one to distinguish between maxima, minima, and inflection points.
ð3:46Þ
The quantity P(r) ¼ R(r)2r2 is the radial distribution function: when multiplied by dr it gives the probability that the electron will be found between r and r þ dr.6 For an orbital with n ¼ 1 and l ¼ 0, it follows from Table 3.2 that 3 Z PðrÞ ¼ 4 r2 e2Zr=a0 a0 This function is illustrated in Fig. 3.20. Note that it is zero at r ¼ 0 (on account of the factor r2) and approaches zero as r ! 1 on account of the exponential factor. By differentiation with respect to r and setting dP/dr ¼ 0 it is easy to show that P goes through a maximum at a0 ð3:47Þ rmax ¼ Z For a hydrogen atom (Z ¼ 1), rmax ¼ a0. Therefore, the radius that Bohr calculated for the state of lowest energy in a hydrogen atom in his early prequantum mechanical model of the atom is in fact the most probable distance of the electron from the nucleus in the quantum mechanical model. Note that this most probable radius decreases in hydrogenic atoms as Z increases, because the electron is drawn closer to the nucleus as the charge of the latter increases. .......................................................................................................
6. For an l ¼ 0 wavefunction (an s-orbital), R2r2 is equivalent to 4pr2jcj2.
3.13 ATOMIC ORBITALS
j
91
z
3.13 Atomic orbitals
y
x (a) z
x
y
(b) Fig. 3.21 Two representations
of the probability density corresponding to a 1s-orbital: (a) the density represented by the darkness of shading, (b) the boundary surface of the orbital.
One-electron wavefunctions in atoms are called atomic orbitals; this name was chosen because it conveys a sense of less certainty than the term ‘orbit’ of classical theory. For historical reasons, atomic orbitals with l ¼ 0 are called s-orbitals, those with l ¼ 1 are called p-orbitals, those with l ¼ 2 are called d-orbitals, and those with l ¼ 3 are called f-orbitals. When an electron is described by the wavefunction cnlml we say that the electron occupies the orbital. An electron that occupies an s-orbital is called an s-electron, and similarly for electrons that occupy other types of orbitals. The shapes of atomic orbitals can be expressed in a number of ways. One way is to denote the probability of finding an electron in a region by the density of shading there (Fig. 3.21). A simpler and generally adequate procedure is to draw the boundary surface, the surface of constant probability within which there is a specified proportion of the probability density (typically 90 per cent). For real forms of the orbitals, the sign of the wavefunction itself is often indicated either by tinting the positive amplitude part of the boundary surface or by attaching þ and signs to the relevant lobes of the orbitals. There are few occasions when a precise portrayal of either the amplitude or the probability density is required, and the qualitative boundary surfaces shown in Fig. 3.22 are generally adequate. The boundary surfaces in Fig. 3.22 show that s-orbitals are spherically symmetrical as Y00 is a constant independent of angle; we have also already seen that s-orbitals differ from other types of orbitals insofar as they have a non-zero amplitude at the nucleus. This feature stems from their lack of orbital angular momentum. It may be puzzling why, with no orbital angular momentum, an s-orbital can exist, because a classical electron without angular momentum would plunge into the nucleus as a result of the nuclear attraction. The answer is found in a quantum mechanical competition between kinetic and potential energies. For an s-electron to cluster close to the nucleus and hence minimize its potential energy, it needs a wavefunction that peaks strongly at the nucleus and is zero elsewhere. However, such a wavefunction is sharply curved, and, on account of its high curvature, corresponds to a very high kinetic energy for the electron. If, instead, the wavefunction spreads over a very wide region with a gentle curvature, then although its kinetic energy will be low, its potential energy will be high because it spends so much time far from the nucleus. The lowest total energy is obtained when the wavefunction is a compromise between confined-but-curved and dispersed-butgently-curved. The three p-orbitals with a given value of n correspond to the three values that ml may have, namely 0 and 1. The orbital with ml ¼ 0 is real (see Y10 in Table 3.1) and has zero component of angular momentum around the z-axis; it is called a pz-orbital. The other two orbitals, pþ and p, are complex, and have their maximum amplitude in the xy-plane
92
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM z
z
z
y
y x
x
y x
ppyz
pz
ppxz
dx 2 – y2 dz 2
Fig. 3.22 Boundary surfaces for p-
dxy
dzx
dyz
and d-orbitals.
(recall Fig. 3.14): 1=2 3 pz ¼ Rn1 ðrÞ cos y 4p 1=2 3 pþ ¼ Rn1 ðrÞ sin y eif 8p 1=2 3 p ¼ Rn1 ðrÞ sin y eif 8p
ð3:48Þ
It is usual to depict the real and imaginary components, and to call these orbitals px and py: 1=2 1 3 px ¼ pffiffiffi ðp pþ Þ ¼ Rn1 ðrÞ sin y cos f 4p 2 ð3:49Þ 1=2 i 3 py ¼ pffiffiffi ðp þ pþ Þ ¼ Rn1 ðrÞ sin y sin f 4p 2 The complex orbitals are the appropriate forms to use in atoms and linear molecules where there are no well-defined x- and y-axes; the real forms are more appropriate when x- and y-axes are well defined, such as in non-linear
3.13 ATOMIC ORBITALS
j
93
molecules. All three real orbitals (px, py, and pz) have the same double-lobed shape, but aligned along the x-, y-, and z-axes, respectively. Example 3.5 How to analyse the probability distribution of an electron
What is the most probable point in space at which a hydrogenic 2pz-electron will be found, and what is the probability of finding the electron inside a sphere of radius R centred on the nucleus? Method. For the first part, we need to inspect the form of the wavefunction and identify the location of the maximum amplitude by considering the maxima in r, y, and f separately. The wavefunction itself is given by combining the information in Tables 3.1 and 3.2, and using n ¼ 2, l ¼ 1, and ml ¼ 0. For the second part, we integrate j c j 2 over a sphere of radius R (that is, over all angles and over all distances between 0 and R). Answer. The wavefunction we require is c210 ¼ R21Y10. The spherical harmonic is proportional to cos y, and its maximum amplitude therefore lies at y ¼ 0 or p, which is along the z-axis. The wavefunction is constant with respect to the azimuth f. The radial wavefunction is proportional to rer/2 with r ¼ (Z/a0)r. To find the location of the maximum of this function we differentiate with respect to r (which is proportional to r) and set the result equal to zero: d r=2 r re ¼ 1 er=2 ¼ 0 dr 2
It follows that the maximum occurs at r ¼ 2, or at r ¼ 2a0/Z. There are two points at which the probability reaches a maximum, at r ¼ 2, y ¼ 0 on the positive z-axis and at r ¼ 2, y ¼ p on the negative z-axis. For the second part of the question, we need to integrate: Z Z R PðrÞ ¼ R221 jY10 j2 dt ¼ R221 r2 dr Sphere of radius R
0
We have used the fact that the spherical harmonics are normalized to 1 when integrated over the surface of a sphere. It then follows from Table 3.2 that Z 1 Z 3 R 2 r 2 PðrÞ ¼ r e r dr 24 a0 0 with r ¼ (Z/a0)r. Therefore, Z Z 1 Z 5 R 4 Zr=a0 1 ZR=a0 4 x PðrÞ ¼ r e dr ¼ x e dx 24 a0 24 0 0 ( ) ZR 1 ZR 2 1 ZR 3 1 ZR 4 ZR=a0 þ þ þ e ¼1 1þ a0 2 a0 6 a0 24 a0 For a hydrogen atom with Z ¼ 1, we find that the probability of the electron being within a sphere of radius 2a0 is Pð2a0 Þ ¼ 1 7e2 ¼ 0:053 Self-test 3.5. Repeat the calculation for a 2s-electron in a hydrogenic atom and evaluate P(2a0) for a hydrogen atom.
94
j
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM z
l = 3, ml = 0 z
l = 3, ml = ±1 z
l = 3, ml = ±2 z
l = 3, ml = ±3
Fig. 3.23 The real parts of the
wavefunctions for the seven atomic orbitals with l ¼ 3. Note that depicted in this way the unique form of the wavefunction with ml ¼ 0 is seen to be a part of a family of cylindrically symmetrical functions.
To derive the expressions for the d-orbitals, we have also used the trigonometric relations cos 2f ¼ cos2f sin2f and sin 2f ¼ 2 sin f cos f.
There are five d-orbitals (l ¼ 2) for n 3. All except the orbital with ml ¼ 0 are complex, and correspond to definite states of orbital angular momentum around the z-axis. These complex orbitals have cylindrical symmetry around the z-axis; however, it is more common to display them as their real components, as in Fig. 3.22: 5 1=2 5 1=2 2 Rn2 ðrÞð3cos y 1Þ ¼ Rn2 ðrÞð3z2 r2 Þ=r2 dz2 ¼ d0 ¼ 16p 16p 1 15 1=2 dx2 y2 ¼ pffiffiffi ðdþ2 þ d2 Þ ¼ Rn2 ðrÞðx2 y2 Þ=r2 16p 2 1=2 1 15 dxy ¼ pffiffiffi ðdþ2 d2 Þ ¼ Rn2 ðrÞxy=r2 4p i 2 1=2 1 15 dyz ¼ pffiffiffi ðdþ1 þ d1 Þ ¼ Rn2 ðrÞyz=r2 4p i 2 1=2 1 15 dzx ¼ pffiffiffi ðdþ1 d1 Þ ¼ Rn2 ðrÞzx=r2 4p 2 The notation stems from the identification of the angular dependence of the orbitals with the relations x ¼ r sin y cos f and so on (eqn 3.17). In deriving these results, we have used the phases of the spherical harmonics specified in Table 3.1. The shapes of f-orbitals (l ¼ 3) are shown in Fig. 3.23. Once the wavefunctions of orbitals are available it is a simple matter to calculate various properties of the electron distributions they represent. For example, the mean radius of an orbital can be evaluated by calculating the expectation value of r by using one of the radial wavefunctions given in Table 3.2. However, it is usually much easier to use one of the following general expressions that are obtained by using the general properties of associated Laguerre functions to evaluate the expectation values: n2 a0 lðl þ 1Þ 1 þ 12 1 hrinlml ¼ n2 Z ð3:50Þ 1 Z ¼ r nlml a0 n2 Note that the first of these expressions shows that the mean radius of an ns-orbital is greater than that of an np-orbital, which is contrary to what one might expect on the basis of the centrifugal effect of orbital angular momentum. It is due to the existence of an additional radial node in the ns-orbital, which tends to extend its radial distribution function out to greater distances. The fact that the average value of 1/r is independent of l is in line with the degeneracy of hydrogenic atoms, for the Coulomb potential energy of the electron is proportional to the mean value of 1/r, and the result implies that all orbitals of a given shell have the same potential energy.
3.14 The degeneracy of hydrogenic atoms We have already seen that the energies of hydrogenic orbitals depend only on the principal quantum number n. To appreciate this conclusion, we can note
3.14 THE DEGENERACY OF HYDROGENIC ATOMS
j
95
that the virial theorem (Section 2.17) for a system in which the potential is Coulombic (s ¼ 1) implies that hEK i ¼ 12 hEP i
ð3:51Þ
However, we have just seen that the mean value of 1/r is independent of l; therefore both the average potential energy and (by the virial theorem) the average kinetic energy are independent of l. Hence the total energy is independent of l, and all orbitals of a given shell have the same energy. Because the permissible values of l are l ¼ 0, 1, . . . , n 1, and for each value of l there are 2l þ 1 orbitals, the degeneracy of a level with quantum number n is gn ¼
n1 X
ð2l þ 1Þ ¼ n2
ð3:52Þ
l¼0
(a)
(b)
(c)
Fig. 3.24 A representation of
the origin of the degeneracy of 2s- and 2p-orbitals in hydrogenic atoms. The object (a) can be rotated into (b), corresponding (when the projection on the two-dimensional plane is inspected) to the rotation of a 2py-orbital into a 2px-orbital. However, rotation about another axis results in a projection that corresponds to a 2s-orbital (c). Thus, in a space of higher dimension, rotations can interconvert 2s- and 2p-orbitals.
as alluded to in Section 3.11. The degeneracy of orbitals with the same value of n but different l is unique to hydrogenic atoms and is lost in the presence of more than one electron. However, the degeneracy of the orbitals with different values of ml but the same values of n and l remains even in the presence of many electrons7 because orbitals with different ml differ only in the orientation of their angular momentum relative to an arbitrary axis. The high degeneracy of a hydrogenic atom is an example of an accidental degeneracy, because there is no obvious rotation of the atom that allows an s-orbital to be transformed into a p-orbital, or some other orbital (recall Section 2.15). However, the Coulomb potential does have a hidden symmetry, a symmetry that is not immediately apparent. This hidden symmetry shows up in spaces of dimension higher than 3. It implies that a four-dimensional being would be able to see at a glance that a 2s-orbital can be rotated into a 2p-orbital, and would therefore not be surprised at their degeneracy, any more than we are surprised at the degeneracy of the three 2p-orbitals. A way of illustrating this hidden symmetry is shown in Fig. 3.24, where we have imagined how a two-dimensional being might experience the projection of a patterned sphere. It is quite easy for us to see that one of the rotations of the sphere results in a change in the projection of the sphere which would lead a Flatlander to think that a p-orbital has been transformed into an s-orbital. We, in our three dimensions, can easily see that the orbitals are related by symmetry; the lowdimensional being, however, might not, and would remain puzzled about the degeneracy. The hydrogen atom has exactly the same kind of higherdimensional symmetry.
.......................................................................................................
7. The degeneracy of states with different values of ml can be removed by the presence of an external electric or magnetic field (Sections 7.19--7.21).
j
96
3 ROTATIONAL MOTION AND THE HYDROGEN ATOM
PROBLEMS 3.1 The rotation of the HI molecule can be pictured as an orbiting of the hydrogen atom at a radius of 160 pm about a virtually stationary I atom. If the rotation is thought of as taking place in a plane (a restriction removed later in Problem 3.14), what are the rotational energy levels? What wavelength of radiation is emitted in the transition ml ¼ þ1 ! ml ¼ 0? 3.2 Confirm eqn 3.2 for the laplacian in two dimensions. 3.3 Show that lz ¼ (h/i)q/qf (that is, confirm eqn 3.9) for a particle confined to a planar surface. 3.4 Show that the wavefunctions in eqn 3.11 are mutually orthogonal. 3.5 Calculate the rotational energy levels of a compact disk of radius 10 cm, mass 50 g free to rotate in a plane. To what value of ml does a rotation rate of 100 Hz correspond? 3.6 Construct the analogues of Figs 3.4 and 3.5 for the states of a rotor with ml ¼ þ3 and þ4. P iml f 3.7 (a) Construct a wavepacket C ¼ N 1 ml ¼0 ð1=ml !Þe and normalize it to unity. Sketch the form of j C j 2 for 0 f 2p. (b) Calculate hfi, hsin fi, and hlzi. (c) Why is hlzi h? Hint. DrawP on a variety of pieces of n x information, including 1 n¼0 x =n! ¼ e and the following integrals: Z
2p 0
ez cos f df ¼ 2pI0 ðzÞ
Z
2p
cos f ez cos f df ¼ 2pI1 ðzÞ
0
with I0(2) ¼ 2.280 . . . , I1(2) ¼ 1.591 . . . ; the I(z) are modified Bessel functions. 3.8 Investigate the properties of the more general P ml wavepacket C ¼ N 1 ða =ml !Þeiml f and show that ml ¼0 when a is large hlzi ah. Hint. Proceed as in the last problem. The large-value expansions of I0(z) and I1(z) are I0(z) ’ I1(z) ’ ez/(2pz)1/2. 3.9 Confirm that the wavefunctions for a particle on a sphere may be written c(y,f) ¼ Y(y)F(f) by the method of separation of variables, and find the equation for Y.
volume element for the integration is siny dy df, with 0 f 2p and 0 y p. 3.13 (a) Confirm that the radius of gyration of a solid uniform sphere of radius R is r ¼ (25)1/2R. (b) What is the radius of gyration of a solid uniform cylinder of radius R and length l? 3.14 Modify Problem 3.1 so that the molecule is free to rotate in three dimensions. Calculate the energies and degeneracies of the lowest three rotational levels, and predict the wavelength of radiation emitted in the l ¼ 1 ! 0 transition. In which region of the electromagnetic spectrum does this wavelength appear? 3.15 Calculate the angle that the angular momentum vector makes with the z-axis when the system is described by the wavefunction clml. Show that the minimum angle approaches zero as l approaches infinity. Calculate the allowed angles when l is 1, 2, and 3. 3.16 Draw the analogues of Fig. 3.23 for l ¼ 2. Observe how the maxima of j Y j 2 migrate into the equatorial plane as j ml j increases, and relate the diagrams to the conclusions drawn in Problem 3.15. 3.17 Calculate the mean kinetic and potential energies of an electron in the ground state of the hydrogen atom, and confirm that the virial is satisfied. R theorem 2 2 Hint. Evaluate hTi ¼ ( h /2m) c r c dt and hVi ¼ 1s 1s R (e2/4pe0) c1s (1/r)c1sdt. The laplacian is given in eqn 3.18 and the virial theorem is dealt with in Further information 3. 3.18 Confirm that the radial wavefunctions u10, u20, and u31 satisfy the radial wave equation, eqn 3.40. Use Table 3.2. 3.19 Locate the radial nodes of the (a) 2s-orbital, (b) 3s-orbital of the hydrogen atom. 3.20 Calculate (a) the mean radius, (b) the mean square radius, and (c) the most probable radius of the 1s-, 2s-, and 3s-orbitals of a hydrogenic atom of atomic number Z. Hint. For the most probable radius look for the principal maximum of the radial distribution function.
3.10 Confirm eqn 3.18 for the laplacian in three dimensions.
3.21 Calculate the probability of finding an electron within a sphere of radius a0 for (a) a 3s-orbital, (b) a 3p-orbital of the hydrogen atom.
3.11 Confirm that the Schro¨dinger equation for a particle free to rotate in three dimensions does indeed separate into equations for the variation with y and f.
3.22 Calculate the values of (a) hri and (b) h1/ri for a 3s- and a 3p-orbital.
3.12 (a) Confirm that Y1,þ1 and Y2,0 as listed in Table 3.1 are solutions of the Schro¨dinger equation for a particle on a sphere. (b) Confirm by explicit integration that Y1, þ 1 and Y2,0 are normalized and mutually orthogonal. Hint. The
3.23 Confirm that c1s and c2s are mutually orthogonal. 3.24 A quantity important in some branches of spectroscopy (Section 13.16) is the probability of an electron being found at the same location as the nucleus.
PROBLEMS
Evaluate this probability density for an electron in the 1s-, 2s-, and 3s-orbitals of a hydrogenic atom. 3.25 Another quantity of interest in spectroscopy is the average value of l/r3 (for example, the average magnetic dipole interaction between the electron and nuclear magnetic moments depends on it). Evaluate h1/r3i for an electron in a 2p-orbital of a hydrogenic atom. 3.26 Calculate the difference in ionization energies of 1H and 2H on the basis of differences in their reduced masses. 3.27 For a given principal quantum number n, l takes the values 0, 1, . . . , n 1 and for each l, ml takes the values l, l 1, . . . , l. Confirm that the degeneracy of the term with principal quantum number n is equal to n2 in a hydrogenic atom. 3.28 Confirm, by drawing pictures like those in Fig. 3.24, that a whimsical Flatlander might be shown that 3s-, 3p-, and 3d-orbitals are degenerate. 3.29 The state of the electron in a He þ ion is described by the wavefunction: c(r,y,f) ¼ R41(r)Y11(y,f). Determine (a) the energy of the electron; (b) the magnitude of the angular momentum vector of the electron; and (c) the projection of the angular momentum vector on to the z-axis. In addition, draw as complete a picture as possible of the vector model of the electron angular momentum. In your picture, specify as many of the lengths and angles as possible. Hint. For the last part of this problem, you need not be concerned with the radial component of c.
j
97
3.30 A diatomic molecule of reduced mass 2.000 10 26 kg and fixed bond length 250.0 pm is rotating about its centre of mass in the xy plane. The state of the molecule is described by the normalized wavefunction c(f). When the total angular momentum of different molecules is measured, two possible results are obtained: a value of 3 h for 25 per cent of the time and a value of 3 h for 75 per cent of the time. However, when the rotational energy of the molecules is measured, the result is surprising. (a) What is the expectation value of the angular momentum? (b) Write down an expression for the normalized wavefunction c(f). (c) What is the result of measuring the energy? Explain (briefly) why you, with your knowledge of quantum mechanics, are not surprised by what is found. 3.31 The state of the electron in a Li2þ ion is described by the normalized wavefunction cðr, y, fÞ ¼ ð13Þ1=2 R42 ðrÞY2;1 ðy, fÞ þ 23iR32 ðrÞY2;1 ðy, fÞ ð29Þ1=2 R10 ðrÞY0;0 ðy, fÞ
(a) If the total energy of different Li2 þ ions in this state is measured, what values will be found? (b) If more than one value is found, what is the probability of obtaining each result and what is the average value? (c) If the magnitude of the total angular momentum is measured, what values will be found? (d) If more than one value is possible, what is the probability of obtaining each result and what is the average value?
4 The angular momentum operators 4.1 The operators and their commutation relations 4.2 Angular momentum observables 4.3 The shift operators The definition of the states 4.4 The effect of the shift operators 4.5 The eigenvalues of the angular momentum 4.6 The matrix elements of the angular momentum 4.7 The angular momentum eigenfunctions 4.8 Spin The angular momenta of composite systems 4.9 The specification of coupled states
Angular momentum
In this chapter, we develop the material introduced in Chapter 3 by showing that many of the results obtained there can be inferred from the properties of operators, as introduced in Chapter 1. For instance, although we have seen that solving the Schro¨dinger equation leads to the conclusion that orbital angular momentum is quantized, the same conclusion can in fact be reached from the angular momentum operators directly without solving the Schro¨dinger equation explicitly. A further point is that because the development in this chapter will be based solely on the commutation properties of the angular momentum operators, it follows that the same conclusions apply to observables that are described by operators with the same commutation properties. Therefore, whenever we meet a set of operators with the angular momentum commutation rules, we will immediately know all the properties of the corresponding observables. This generality is one of the reasons why angular momentum is of such central importance in quantum mechanics. Angular momentum has many more mundane applications. It is central to the discussion of the structures of atoms (we have already caught a glimpse of that in the discussion of hydrogenic atoms), to the discussion of the rotation of molecules, as well as to virtually all forms of spectroscopy. We shall draw heavily on this material when we turn to the applications of quantum mechanics in Chapter 7 onwards.
4.10 The permitted values of the total angular momentum 4.11 The vector model of coupled angular momenta 4.12 The relation between schemes 4.13 The coupling of several angular momenta
The angular momentum operators It follows from the general introduction to quantum mechanics in Chapter 1, that the quantum mechanical operators for angular momentum can be constructed by replacing the position, q, and linear momentum, pq, variables in the classical definition of angular momentum by operators that satisfy the commutation relation ½q, pq0 ¼ ihdqq0
ð4:1Þ
We shall set up these angular momentum operators and then show how to determine their commutation relations.
4.1 THE OPERATORS AND THEIR COMMUTATION RELATIONS
j
99
4.1 The operators and their commutation relations l
r
p
In classical mechanics, the angular momentum, l, of a particle travelling with linear momentum p at an instantaneous position r on its path is defined as the vector product l ¼ r p (Fig. 4.1). Note that l displays the sense of rotation according to the right-hand screw rule: it points in the direction a right-hand (conventional) screw travels when it is turned in the same sense as the rotation. If the position of the particle is expressed in terms of the components of the vector r ¼ xi þ yj þ zk
Fig. 4.1 The definition of orbital
angular momentum as l ¼ r p. Note that the angular momentum vector l stands perpendicular to the plane of the motion of the particle.
where i, j, and k are mutually orthogonal unit vectors, and the linear momentum is expressed in terms of its components, p ¼ px i þ py j þ pz k then it follows that the angular momentum can be expressed in terms of its components l ¼ lx i þ ly j þ lz k as l ¼ r p ¼ ðypz zpy Þi þ ðzpx xpz Þj þ ðxpy ypx Þk
ð4:2Þ
We can therefore identify the three components of the angular momentum as lx ¼ ypz zpy
ly ¼ zpx xpz
lz ¼ xpy ypx
ð4:3Þ
Note how each component can be generated from its predecessor by cyclic permutation of x, y, and z. The expression for lz matches that given by eqn 3.8. The magnitude, l, of the angular momentum is related to its components by the normal expression for constructing the magnitude of a vector: l2 ¼ lx2 þ ly2 þ lz2
ð4:4Þ
Classical mechanics puts no constraints on the magnitude of angular momentum, which is consistent with the kinetic energy of rotation E ¼ l2/2I being continuously variable too. Nor does it put any constraints on the components of angular momentum about the three axes, other than the requirement, to be consistent with eqn 4.4, that none of the components exceeds the magnitude (jlqj l). The definitions of the components and the magnitude carry over into quantum mechanics, with the q and pq in the definitions of the lq interpreted as operators. The operators lq in the position representation are obtained, as h/i)q/qq: explained in Section 1.5, by replacing q by q and pq by ( h q q h q q h q q ly ¼ lz ¼ ð4:5Þ lx ¼ y z z x x y i qz qy i qx qz i qy qx However, instead of developing the properties of angular momentum in a specific representation, it is more general, more powerful, and more timesaving to develop them without selecting a representation. Later in the chapter we shall make use of the fact that because the operators lq and l2
100
j
4 ANGULAR MOMENTUM
correspond to observables, they must be hermitian (Section 1.8). The property of hermiticity can be demonstrated explicitly in the position representation (see Example 1.5); but it must be true in any representation if the operators are to stand for observables. To make progress, we need to establish the commutation relations of the lq operators. Consider first the commutator of lx and ly: ½lx , ly ¼ ½ypz zpy , zpx xpz ¼ ½ypz , zpx ½ypz , xpz ½zpy , zpx þ ½zpy , xpz ¼ y½pz , zpx 0 0 þ xpy ½z, pz ¼ ihð ypx þ xpy Þ ¼ ihlz
ð4:6Þ
In line 1 we have inserted the definitions. In line 2 we have expanded the commutators term by term. In line 3 we have used the fact that y and px commute with each other and also with z and pz. The same is true of x and py. The remaining commutators can be derived in the same way, but it is more efficient to note that because the three operators lq are obtained from one another by cyclic permutation, then the commutators can be obtained in the same way. We therefore conclude that ½lx , ly ¼ ihlz
½ly , lz ¼ ihlx
½lz , lx ¼ ihly
ð4:7Þ
2
The remaining operator is l , the operator corresponding to the square of the magnitude of the angular momentum. We need its commutator with the operators lq, and proceed as follows. First, we write ½l2 , lz ¼ ½lx2 þ ly2 þ lz2 , lz ¼ ½lx2 , lz þ ½ly2 , lz We have used the fact that the commutator of lz2 and lz is zero: ½lz2 , lz ¼ lz2 lz lz lz2 ¼ lz3 lz3 ¼ 0 Next, consider the following commutator, which we develop by drawing on the three fundamental relations derived above: ½lx2 , lz ¼ lx lx lz lz lx lx ¼ lx lx lz lx lz lx þ lx lz lx lz lx lx ¼ lx ½lx , lz þ ½lx , lz lx ¼ ihðlx ly þ ly lx Þ Similarly, ½ly2 , lz ¼ ihðlx ly þ ly lx Þ The sum of [lx2 , lz] and [ly2 , lz] is zero, so we can conclude that the commutator of l2 with lz is zero. Moreover, because lx, ly, and lz occur symmetrically in l2, all three operators must commute with l2 if any one of them does. That is, ½l2 , lq ¼ 0
ð4:8Þ
for all q. The commutation relations in eqns 4.7 and 4.8 are the foundations for the entire theory of angular momentum. Whenever we encounter four operators
4.3 THE SHIFT OPERATORS
j
101
having these commutation relations, we know that the properties of the observables they represent are identical to the properties we are about to derive. Therefore, we shall say that an observable is an angular momentum if its operators satisfy these commutation relations.1
z
4.2 Angular momentum observables Eigenvalue of lz
Square-root of the eigenvalue of l 2
Fig. 4.2 The cone used to represent a state of angular momentum with specified magnitude and z-component.
We saw in Section 1.16 that observables are complementary and restricted by the uncertainty relation if their operators do not commute, and we have just seen that lz does not commute with either lx or ly. Therefore, although we can specify any one of these components, we cannot specify more than one. However, l2 does commute with all three components, so the magnitude of the angular momentum may be specified simultaneously with any of its components. These conclusions are the quantum mechanical basis of the ‘vector model’ of angular momentum introduced in Section 3.6, where we represent an angular momentum state by a vector of indeterminate orientation on a cone of given side (the magnitude of the momentum) and height (the eigenvalue of lz, Fig. 4.2). At this point, though, we can begin to see that the vector model must be regarded with caution. The commutation relations in eqn 4.7 can be written in a compact fashion as follows: l l ¼ ihl
ð4:9Þ
To confirm this relation, write the left-hand side as a determinant and expand it; then compare it term-by-term with the expression on the right-hand side: this procedure reproduces the three commutation relations (see Problem 4.4). However, it is an elementary feature of vector algebra that the vector product of a vector with itself is zero (the magnitude of a b is proportional to sin y, where y is the angle between the vectors; but when the two vectors are identical that angle is zero). Therefore, because the vector product of l with itself is not zero, we have to conclude that l is not a vector. The vector model is useful only if we realize that it is not the whole truth, and note that l is a vector operator, not a classical vector.
4.3 The shift operators It will prove expedient to introduce linear combinations of the angular momentum operators, called the shift operators. These operators will prove to be particularly useful for establishing the properties of angular momentum and for the evaluation of matrix elements of angular momentum operators. .......................................................................................................
1. Because all the properties of the observables are the same, this seems to be an appropriate course of action. However, the procedure does capture some strange bed-fellows. The electric charge of fundamental particles is described by operators that satisfy the same set of communication relations, but should we regard it—or imagine it—as an angular momentum? Electron spin is also described by the same set of communication relations, but should we regard it—or imagine it—as an angular momentum?
102
j
4 ANGULAR MOMENTUM
One operator, lþ , is called the raising operator; the other, l , is called the lowering operator. They are defined as follows: ð4:10Þ lþ ¼ lx þ ily l ¼ lx ily The inverse relations are lþ þ l lþ l ly ¼ ð4:11Þ lx ¼ 2 2i We shall require the commutators of the shift operators. They are easily derived from the fundamental commutation relations. For example, ½lz , lþ ¼ ½lz , lx þ i½lz , ly ¼ ihly þ hlx ¼ hlþ The other commutation relations are obtained similarly, and all three are ½lz , lþ ¼ hlþ ½lz , l ¼ hl ½lþ , l ¼ 2 hlz ð4:12Þ Furthermore, because l2 commutes with each of its components, it also commutes with l . Therefore, we can add to these relations the rule ½l2 , l ¼ 0
ð4:13Þ
The definition of the states The next task is to see how the commutation relations govern the values of the permitted eigenvalues of l2 and any one of the components lq. It is conventional to call the selected component lz, but that is entirely arbitrary (as is the choice of the direction denoted z). In the course of this development we shall discover that the solutions found in Chapter 3 are incomplete in a very important respect. We shall also set up an elegant way of constructing the spherical harmonics, and find a simple way of evaluating the matrix elements of angular momentum operators.
4.4 The effect of the shift operators We shall suppose that the simultaneous eigenstates of l2 and lz are distinguished by two quantum numbers, which for the time being we shall denote l and ml. The eigenstates are therefore denoted jl, mli. We define ml through the relation lz jl, ml i ¼ ml hjl, ml i ð4:14Þ This relation must be true, because h has the same dimensions as an angular momentum (M L2 T 1), so the eigenvalue of lz must be a numerical multiple of h; we are not presupposing that ml is restricted to discrete values, but that will emerge in due course. All we know is that ml is a real number: that follows from the hermiticity of lz. Because l2 commutes with lz, the state jl, mli is also an eigenstate of l2. At this stage we shall allow for the possibility that the eigenvalues of l2 depend on both quantum numbers, and write l2 jl, ml i ¼ f ðl, ml Þ h2 jl, ml i
ð4:15Þ
4.4 THE EFFECT OF THE SHIFT OPERATORS
j
103
where f is a function that we need to determine: from the work we did in Chapter 3 we know that it will turn out to be equal to l(l þ 1) where l is the maximum value of jmlj, but that is something we shall derive. All we know at this stage is that because l2 is hermitian, f is real. Moreover, because l2 is the sum of squares of hermitian operators, we also know (recall Example 1.9) that its eigenvalues are non-negative. Because l2 lz2 ¼ lx2 þ ly2 , it follows that the eigenvalues of the operator 2 l lz 2 are non-negative: ðl2 lz2 Þjl, ml i ¼ ðlx2 þ ly2 Þjl, ml i 0 However, we also know from the definitions of the effects of l2 and lz2 that h2 jl, ml i ðl2 lz2 Þjl, ml i ¼ ff ðl, ml Þ m2l g
Recall that if j!i is an eigenstate of O, then O2j!i ¼ O!j!i ¼ !Oj!i ¼ !2j!i.
For these two relations to be consistent, it follows that f ðl, ml Þ m2l
ð4:16Þ
To take the next step we use the commutation relations to establish the effect of the shift operators (and see why they are so-called). Consider the effect of the operator lþ on jl, mli. Because jl, mli is an eigenstate of neither lx nor ly, when lþ acts on it, it generates a new state. First, we show that lþ jl,mli is an eigenstate of l2 with the same value of f; that is jl,mli and lþ jl,mli share the same eigenvalue of l2. To do so, consider the effect of l2 on the state obtained by acting with lþ : h2 jl, ml i ¼ f ðl, ml Þ h2 lþ jl, ml i l2 lþ jl, ml i ¼ lþ l2 jl, ml i ¼ lþ f ðl, ml Þ where the first equality follows from the fact that l2 and lþ commute. It follows, because the eigenvalue of l2 for the state lþ jl, mli is the same as that for the original state jl, mli, that lþ leaves the magnitude of the angular momentum unchanged when it acts. Now consider the same argument applied to jl, mli treated as an eigenstate of lz. The conclusion will be different, because lþ and lz do not commute. Instead, we must use the following string of equalities to find the effect of lz on lþ jl,mli: lz lþ jl, ml i ¼ ðlþ lz þ ½lz , lþ Þjl, ml i ¼ ðlþ lz þ hlþ Þjl, ml i hþ hlþ Þjl, ml i ¼ ðml þ 1Þ hlþ jl, ml i ¼ ðlþ ml However, we know from eqn 4.14 that ml +1
lz jl, ml þ 1i ¼ ðml þ 1Þ hjl, ml þ 1i l+
ml
ml –1
Fig. 4.3 The effect of the shift operators l þ and l .
l–
Therefore, the state lþ jl, mli must be proportional to the state jl, ml þ 1i and we can write hjl, ml þ 1i lþ jl, ml i ¼ cþ ðl, ml Þ
ð4:17aÞ
where cþ (l, ml) is a dimensionless numerical coefficient which in due course we shall need to find. We now see why lþ is called a raising operator: when h, it generates from it a state it operates on a state with z-component ml with the same magnitude of angular momentum but with a z-component h (Fig. 4.3). In exactly the same way, the effect one unit greater, (ml þ 1)
104
j
4 ANGULAR MOMENTUM
of the operator l can be shown to lower the z-component from ml h to h: (ml 1) l jl, ml i ¼ c ðl, ml Þ hjl, ml 1i
ð4:17bÞ
where c (l, ml) is another dimensionless numerical coefficient.
4.5 The eigenvalues of the angular momentum The shift operators step ml by 1 each time they operate. However, we have already established from the hermiticity of the operators that m2l cannot exceed f(l, ml); it follows that ml must have a maximum value, which we shall denote l. When we operate with lþ on a state in which ml ¼ l, we generate nothing, because there is no state with a larger value of ml: lþ jl, li ¼ 0 This relation will give us the value of the unknown function f. When acted on by l , it gives l lþ jl, li ¼ 0 However, the product l lþ can be expanded as follows: l lþ ¼ ðlx ily Þðlx þ ily Þ ¼ lx2 þ ly2 þ ilx ly ily lx ¼ lx2 þ ly2 þ i½lx , ly ¼ l2 lz2 þ iðihlz Þ
ð4:18Þ
Therefore, the last equation can be written ðl2 lz2 hlz Þjl, li ¼ 0 When we rearrange this expression and use the definition of the effect of lz on a state, we obtain hlz Þjl, li ¼ ðl2 þ lÞ h2 jl, li l2 jl, li ¼ ðlz2 þ It follows that f ðl, lÞ ¼ lðl þ 1Þ We have already established that when l acts on a state, it leaves the eigenvalue of l2 unchanged. Therefore, all the states jl, li, jl, l 1i, etc. have the same eigenvalue of l2. Therefore, f ðl, ml Þ ¼ lðl þ 1Þ
for ml ¼ l, l 1, . . .
We know that there is a lower bound on ml because the eigenvalue of lz2 cannot exceed the eigenvalue of l2, and for the moment we denote this lower bound by k. It is quite easy to show that k ¼ l. To see that this is the case, we start from l jl,ki ¼ 0, and by a similar argument but using lþ l jl,ki ¼ 0, conclude that f(l,k) ¼ k(k 1). However, because f(l,ml) is independent of ml, we must have l(l þ 1) ¼ k(k 1). Of the two solutions k ¼ l and k ¼ l þ 1, only the former is acceptable (the lower bound must be below the upper bound!). Therefore, f ðl, ml Þ ¼ lðl þ 1Þ
for ml ¼ l, l 1, . . . , l
4.5 THE EIGENVALUES OF THE ANGULAR MOMENTUM
j
105
At this point we can put the spare quantum number l to work, and identify it as l, the maximum value of jmlj. Then, f ðl, ml Þ ¼ lðl þ 1Þ
for ml ¼ l, l 1, . . . , l
ð4:19Þ
That is, we now know that h2 jl, ml i l2 jl, ml i ¼ lðl þ 1Þ
ð4:20Þ
and we see that the value of l (the maximum value of ml) determines the magnitude of the angular momentum. We already know that hjl, ml i lz jl, ml i ¼ ml
ð4:21Þ
and so we have an effectively complete description of angular momentum. Finally, we need to decide on the allowed values of l and ml. As we have seen, the shift operators step the states jl, mli from jl, þ li to jl, li in unit steps. The symmetry of this ladder of states allows for only two types of value for l: it may be integral or half-integral. For example, we can have l ¼ 2, to give the ladder ml ¼ þ2, þ1, 0, 1, 2, or we could have l ¼ 32, to give ml ¼ þ32, þ12, 12, 32. We cannot obtain a symmetrical ladder with any other type of value (l ¼ 34, for instance, would give the unsymmetrical ladder ml ¼ þ34, 14). We can summarize the conclusions so far. On the basis of the hermiticity of the angular momentum operators and their commutation relations, we have shown: 1. The magnitude of the angular momentum is confined to the values h, with l ¼ 0, 12, 1, . . . . {l(l þ 1)}1/2 h 2. The component on an arbitrary z-axis is limited to the 2l þ 1 values ml with ml ¼ l, l 1, . . . , l. These conclusions differ in one detail from those obtained by solving the Schro¨dinger equation in Chapter 3. There we saw that l was confined to the integral values l ¼ 0, 1, 2, . . . . In that analysis, we obtained the permitted values of l by imposing cyclic boundary conditions. What the present analysis does is to show that angular momentum may be described by half-integral quantum numbers, but such quantum numbers do not necessarily apply to a particular physical situation. For orbital angular momentum, where the Born interpretation requires cyclic boundary conditions to be satisfied, only integral values are admissible. Where cyclic boundary conditions are not relevant, as for the intrinsic angular momentum known as spin, the halfintegral values may be appropriate. We shall use the following notation to emphasize that there is a distinction between angular momenta according to the boundary conditions that have to be satisfied. For orbital angular momenta, when the boundary conditions on the wavefunctions allow only integral quantum numbers, we shall use the notation l and ml and write states as jl,mli. When internal angular momentum (spin) is being considered, we shall use the notation s and ms for the (possibly half-integral) quantum numbers and write the states js,msi. When the discussion is general and applicable to either kind of angular momentum, we shall use the quantum numbers j and mj, and write the states as jj, mji. The
106
j
4 ANGULAR MOMENTUM
expressions we have deduced so far may therefore be written in this general notation as h2 jj, mj i jz jj, mj i ¼ mj hjj, mj i j2 jj, mj i ¼ jð j þ 1Þ
ð4:22Þ
with mj ¼ j, j 1, . . . , j.
4.6 The matrix elements of the angular momentum One outstanding problem at this point is the value of the coefficient c introduced in connection with the effect of the shift operators: hjj, mj 1i j jj, mj i ¼ c ð j, mj Þ
ð4:23Þ
Because the states jj,mji form an orthonormal set, multiplication from the left by the bra h j,mj 1j gives h h j, mj 1jj jj, mj i ¼ c ð j, mj Þ
ð4:24Þ
So, we need to know the coefficients if we want to know the values of these matrix elements. Matrix elements of this kind occur in connection with the calculation of magnetic properties and the intensities of transitions in magnetic resonance (Chapter 13). The first step involves finding two expressions for the matrix elements of the operator j jþ . First, we can use eqn 4.18 to write j jþ jj, mj i ¼ ð j2 j2z hjz Þjj, mj i ¼ f jð j þ 1Þ mj ðmj þ 1Þg h2 jj, mj i Alternatively, we can use eqn 4.23 to write hjj, mj þ 1i ¼ cþ ð j, mj Þc ð j, mj þ 1Þ h2 jj, mj i j jþ jj, mj i ¼ j cþ ð j, mj Þ Comparison of the two expressions shows that cþ ð j, mj Þc ð j, mj þ 1Þ ¼ jð j þ 1Þ mj ðmj þ 1Þ
ð4:25Þ
The next step is to find a relation between the two coefficients that occur in the last expression. We shall base the calculation on the matrix element h h j, mj jj jj, mj þ 1i ¼ c ð j, mj þ 1Þ and the hermiticity of jx and jy. Consider the following string of manipulations: h j, mj jj jj, mj þ 1i ¼ h j, mj jjx ijy jj, mj þ 1i ¼ h j, mj jjx jj, mj þ 1i ih j, mj jjy jj, mj þ 1i ¼ h j, mj þ 1jjx jj, mj i ih j, mj þ 1jjy jj, mj i ½by hermiticity ¼ fh j, mj þ 1jjx jj, mj i þ ih j, mj þ 1jjy jj, mj ig ¼ h j, mj þ 1jjþ jj, mj i The relation just derived, which reads h j, mj jj jj, mj þ 1i ¼ h j, mj þ 1jjþ jj, mj i
ð4:26Þ
shows that j and jþ are each other’s hermitian conjugate. Neither operator is hermitian, and so neither operator corresponds to a physical
4.6 THE MATRIX ELEMENTS OF THE ANGULAR MOMENTUM
j
107
observable. In general, two operators A and B are each other’s hermitian conjugate if hajAjbi ¼ hbjBjai
ð4:27Þ
The relation we have just derived implies a relation between the coefficients c . Because the matrix element on the left of eqn 4.26 is equal to h and that on the right is equal to cþ (j,mj) h, it follows that c (j,mj þ 1) c ð j, mj þ 1Þ ¼ cþ ð j, mj Þ
ð4:28Þ
It then follows from eqn 4.25 that jcþ ð j, mj Þj2 ¼ jð j þ 1Þ mj ðmj þ 1Þ If we make a convenient choice of phase (choosing cþ to be real and positive), it follows that cþ ð j, mj Þ ¼ fjð j þ 1Þ mj ðmj þ 1Þg1=2
ð4:29aÞ
Moreover, because c (j,mj) ¼ cþ(j, mj 1), we can also write c ð j, mj Þ ¼ fjð j þ 1Þ mj ðmj 1Þg1=2
ð4:29bÞ
With these matrix elements established, we can calculate a wide range of other quantities, as illustrated in the following example. Example 4.1 How to evaluate matrix elements of the angular momentum
Evaluate the matrix elements (a) hj, mj þ 1jjxjj, mji, (b) h j, mj þ 2jjxjj, mji, and (c) h j, mj þ 2jjx2jj, mji. Method. Because we know the matrix elements of the shift operators,
one approach is to express all the operators in the questions in terms of them and then to use eqns 4.24 and 4.29. Note that j2x ¼ jx jx and h j0 ; m0j jj; mj i ¼ dj0 j dmj0 mj . Answer.
ðaÞ h j, mj þ 1jjx jj, mj i ¼ 12 h j, mj þ 1jjþ þ j jj, mj i ¼ 12 h j, mj þ 1jjþ jj, mj i þ 12 h j, mj þ 1jj jj, mj i h ¼ 12 cþ ð j, mj Þ because h j, mj þ 1jj jj, mji / hj, mj þ 1jj, mj 1i ¼ 0. ðbÞ h j, mj þ 2jjx jj, mj i ¼ 0 because j steps mj only by one unit, and the resulting states are orthogonal to the state jj, mj þ 2i. ðcÞ h j, mj þ 2jj2x jj, mj i ¼ 14 h j, mj þ 2jj2þ þ j2 þ jþ j þ j jþ jj, mj i ¼ 14 h j, mj þ 2jj2þ jj, mj i h2 ¼ 14 cþ ð j, mj þ 1Þcþ ð j, mj Þ ¼ 14 fjð j þ 1Þ ðmj þ 1Þðmj þ 2Þg1=2 f jð j þ 1Þ mj ðmj þ 1Þg1=2 h2
108
j
4 ANGULAR MOMENTUM
Comment. Note that it is quite easy to spot short-cuts, as in (c), where it should be obvious that only j2þ can contribute to the matrix element. Self-test 4.1. Evaluate the matrix element hj,mj þ 1jjx3jj,mji.
4.7 The angular momentum eigenfunctions Now we consider orbital angular momentum explicitly. This version of the general theory refers to the angular momentum arising from the distribution of a particle in space, so it is subject to cyclic boundary conditions on the wavefunctions. As we saw in Chapter 3, these conditions limit the angular momentum quantum numbers to integral values, and we denote them l and ml. In Chapter 3 we saw that the wavefunctions are solutions of a secondorder differential equation, and we asserted (and proved in Further information 9) that they were the spherical harmonics. With the work done in this chapter, we can show that they can also be obtained by solving a first-order differential equation, which is a much simpler task. We begin by finding the wavefunction for the state jl,li (for which ml ¼ l). Once this wavefunction has been determined, the wavefunctions for the states jl,mli can be generated by acting on jl,li with l the appropriate number of times. The equation we have to solve is lþ jl, li ¼ 0 To express this equation as a differential equation, we must adopt a representation for the operators. In the position representation, the orbital angular momentum operators are h q q sin f þ cot y cos f lx ¼ i qy qf h q q cos f cot y sin f ly ¼ i qy qf h q lz ¼ ð4:30Þ i qf These operators are obtained from the cartesian forms given in eqn 4.3 by expressing them in terms of spherical polar coordinates. It follows that the shift operators in the position representation are q q þ i cot y heif lþ ¼ qy qf q q if i cot y ð4:31Þ lþ ¼ he qy qf To obtain these expressions we have used Euler’s relation e ix ¼ cos x i sin x and cot x ¼ 1/tan x ¼ cos x/sin x.
It follows from the equation lþ jl, li ¼ 0 that q if q þ i cot y c ðy, fÞ ¼ 0 he qy qf l;l
4.7 THE ANGULAR MOMENTUM EIGENFUNCTIONS
j
109
This partial differential equation can be separated by writing c(y,f) ¼ Y(y)F(f), for in the normal way (substituting, differentiating, and then dividing through by YF) we then obtain tan y dY i dF ¼ Y dy F df According to the usual separation of variables argument, both sides are equal to a constant, which we denote c. The equation therefore separates into the following two first-order ordinary differential equations: dY dF ¼ cY ¼ ic F tan y dy df The two equations integrate immediately to Y / sinc y
F / eicf
The value of c is found to be l by requiring that lzcl,l ¼ lcl,l. Therefore, the complete solution is cl;l ¼ N sinl y eilf
ð4:32Þ
where N is a normalization constant. This is the explicit form of the spherical harmonic Yll given in Table 3.1, apart from the normalization constant, which can be obtained by integration over the surface of a sphere. With this function found, it is a straightforward matter to apply the operator l to obtain the rest of the functions with a given value of l. Example 4.2 How to construct wavefunctions for states with ml < l
Construct the wavefunction for the state jl, l 1i. Method. We know that l jl, li ¼ c hjl, l 1i. We also know the position
representation form of l (eqn 4.31). We need to combine the two expressions. Answer. In the position representation we have
l cl;l ¼ he if
q q i cot y N sinl yeilf qy qf
¼ Nhe if ðl sinl 1 y cos y iðilÞ cot y sinl yÞeilf ¼ 2Nlh sinl 1 y cos y eiðl 1Þf However, we also know that l jl, li ¼ flðl þ 1Þ lðl 1Þg1=2 hjl, l 1i ¼ ð2lÞ1=2 hjl, l 1i Therefore, cl;l 1 ¼ ð2lÞ1=2 N sinl 1 y cos eiðl 1Þf Comment. If cl,l is normalized to unity, then so is cl,l 1 and all the other states
that can be generated in this way. The normalization constant is 1 ð2l þ 1Þ! 1=2 N¼ l 4p 2 l! Self-test 4.2. Derive an expression for the wavefunction with ml ¼ l 2 in the
same way.
110
j
4 ANGULAR MOMENTUM
4.8 Spin The Dutch physicists George Uhlenbeck and Samuel Goudsmit realized in 1925 that a great simplification of the description of atomic spectra could be obtained if it was assumed that an electron possessed an intrinsic angular momentum with quantum number s ¼ 12 and which could exist in two states with ms ¼ þ12, denoted a or ", and ms ¼ 12, denoted b or #. This intrinsic angular momentum is called the spin of the electron (but footnote 1 of this chapter should be recalled). This realization shed light on a seminal experiment performed several years earlier by Otto Stern and Walther Gerlach. The Stern–Gerlach experiment consisted of preparing a beam of silver atoms and passing them through a strong, inhomogeneous magnetic field. Stern and Gerlach found that the beam was deflected into two directions and ascribed the effect to space quantization and the magnetic moment of the electron. (In an Ag atom, there is a single electron outside a closed shell, so the atom behaves like a single electron on a heavy platform, the rest of the atom.) However, Stern and Gerlach did not realize they had discovered electron spin but rather devised their experiment based on considerations of orbital angular momentum.2 Moreover, although electron spin was discovered in 1925, it appears that it was not until 1927 that the Stern– Gerlach splitting was attributed to the spin of the electron being in either of two directions, to what we would now interpret as the states with ms ¼ þ12 and 12. Spin is a purely quantum mechanical phenomenon in the sense that in a universe in which h ! 0 the spin angular momentum would be zero. Orbital angular momentum survives in a classical world, because l can be allowed h lh can be to approach infinity as h ! 0 and the quantity {l(l þ 1)}1/2 non-zero. Uhlenbeck and Goudsmit’s proposal was initially no more than a hypothesis, but when Dirac showed how to combine quantum mechanics and special relativity, the existence of particles with half-integral angular momentum quantum numbers appeared automatically. The angular momentum operators describe spin, but for s ¼ 12 they do so in a very simple way. If we denote the state j12,þ12i by a and the state j12, 12i by b, then the general expressions given earlier become sz a ¼ þ12 ha sz b ¼ 12 hb s2 a ¼ 34 h2 a
s2 b ¼ 34 h2 b
ð4:33Þ
and the effects of the shift operators are sþ a ¼ 0 sþ b ¼ ha s a ¼ hb s b ¼ 0
ð4:34Þ
It follows that the only non-zero matrix elements of the shift operators are h hajsþ jbi ¼ hbjs jai ¼
ð4:35Þ
.......................................................................................................
2. An enjoyable and amusing account of the Stern–Gerlach experiment and its interpretation can be found in Space Quantization: Otto Stern’s Lucky Star, by B. Friedrich and D. Herschbach, Daedalus, 165, 127 (1998).
4.8 SPIN
Recall that an arbitrary function f can be written as a linear combination of basis set functions {f1, f2, . . . , fn} as P f ¼ n cnfn. The function can be represented as an n 1 column vector 0 1 c1 B c2 C B C f ¼ B .. C @ . A cn The basis set functions themselves can be regarded as column vectors with all components, except one, equal to zero.
j
111
The operators can be written succinctly in terms of matrices by considering their effects on the orthonormal basis set {a, b}: 1 0 a¼ b¼ ð4:36Þ 0 1 With this notation, the effect of the operator sz can be reproduced by the effect of a two-dimensional matrix: 1 0 1 1 1 1 h h ¼ þ12 ha s z a ¼ þ2 ¼ þ2 0 1 0 0 with a similar expression for the effect of sz on b. Likewise, the effect of sx on a, which according to eqns 4.11 and 4.34 is sxa ¼ 12 hb, can be expressed as follows: 0 1 0 1 h h ¼ þ12 hb sx a ¼ þ12 ¼ þ12 1 0 1 0 with a similar expression for the effect of sx on b. In fact, all the properties of the spin-12 operators, including their commutation relations, are reproduced by the matrices: 0 1 0 i 1 0 sy ¼ sz ¼ ð4:37Þ sx ¼ 1 0 i 0 0 1 0 2 0 0 sþ ¼ s ¼ 0 0 2 0 and the relation sq sq ¼ 12 h
ðq ¼ x, y, z, þ , Þ
ð4:38Þ
The set of matrices sx, sy, sz are known collectively as the Pauli matrices. These matrices play an important role in the development of the properties of spin-12 systems, and we shall meet them again. Illustration 4.1 The Pauli representation of commutation relations
To confirm that the Pauli matrices correctly represent the angular momentum commutation relations, we write 0 1 0 i 0 i 0 1 ½sx , sy ¼ 1 0 i 0 i 0 1 0 i 0 i 0 ¼ 0 i 0 i i 0 ¼2 0 i 1 0 ¼ 2i ¼ 2isz 0 1 It then follows from eqn 4.38 that h2 ð2isz Þ ¼ i hsz ½sx , sy ¼ 14 h2 ½sx , sy ¼ 14 as required.
112
j
4 ANGULAR MOMENTUM
The angular momenta of composite systems We now consider a system in which there are two sources of angular momentum, which we denote j1 and j2. The system might be a single particle that possesses both spin and orbital angular momentum, or it might consist of two particles with spin or orbital angular momentum. The question we investigate here is what the commutation rules imply for the total angular momentum j of the system.
4.9 The specification of coupled states The state of particle 1 is fully specified by reporting the quantum numbers j1 and mj1, and the same is true of particle 2 in terms of its quantum numbers j2 and mj2. If we are to be able to specify the overall state as jj1mj1; j2mj2i, we need to know whether all the corresponding operators commute with one another. In fact, operators for independent sources of angular momentum do commute with one another, and we can write ½ j1q , j2q0 ¼ 0
ð4:39Þ
for all the components q ¼ x, y, z and q 0 ¼ x, y, z. One way to confirm this conclusion is to note that in the position representation the operators are expressed in terms of the coordinates and derivatives of each particle separately, and the derivatives for one particle treat the coordinates of the other particle as constants. Operators that refer to independent components of a system always commute with one another. Because the operators j21 and j22 are defined in terms of their components, which commute, so too do these two operators. Hence, all four operators j21 , j1z, j22 , and j2z commute with one another, and it is permissible to express the state as jj1mj1;j2mj2i. We now explore whether the total angular momentum, j ¼ j1 þ j2, can also be specified. First, we investigate whether j is indeed an angular momentum. To do so, we evaluate the commutators of its components, such as ½ jx , jy ¼ ½ j1x þ j2x , j1y þ j2y ¼ ½ j1x , j1y þ ½ j2x , j2y þ ½ j1x , j2y þ ½ j2x , j1y ¼ ihj1z þ ihj2z þ 0 þ 0 ¼ ihjz
ð4:40Þ
This commutation relation, and the other two that can be derived from it by cyclic permutation of the coordinate labels, is characteristic of angular momentum, so j is an angular momentum (j1 j2, on the other hand, is not). Because j is an angular momentum, we can conclude without further work h with j integral or half-integral, and its that its magnitude is {j(j þ 1)}1/2 h with mj ¼ j, j 1, . . . , j. z-component has the values mj We now need to work towards discovering which values of j can exist in the system. The initial question is whether we can actually specify j if j1 and j2 have been specified. Because j21 commutes with all its components and
4.10 THE PERMITTED VALUES OF THE TOTAL ANGULAR MOMENTUM
j
113
j22 commutes with its, and because j2 can be expressed in terms of those same components, it follows that ½ j2 ,j21 ¼ ½ j2 ,j22 ¼ 0
ð4:41Þ
Therefore, we can conclude that the eigenvalues of j21 , j22 , and j2 can be specified simultaneously. For instance, a p-electron (for which l ¼ 1 and s ¼ 12) can be regarded as having a well-defined total angular momentum with a magnitude given by some value of j (the actual permitted values of which we have yet to find). Because j2 commutes with its own components, in particular it commutes with jz ¼ j1z þ j2z. Therefore, we know that we can specify the value of mj as well as j. At this point, we have established that a state of coupled angular momentum can be denoted j j1j2;jmji. Note, however, that we have not yet established that it can be specified more fully as j j1mj1 j2mj2 ;jmji because we have not yet established whether j2 commutes with j1z and j2z. To explore this point we proceed as follows: ½ j1z , j2 ¼ ½ j1z , j2x þ ½ j1z , j2y þ ½ j1z , j2z ¼ ½ j1z , ð j1x þ j2x Þ2 þ ½ j1z , ð j1y þ j2y Þ2 þ ½ j1z ; ð j1z þ j2z Þ2 ¼ ½ j1z , j21x þ 2j1x j2x þ ½ j1z , j21y þ 2j1y j2y ¼ ½ j1z , j21x þ j21y þ 2½ j1z , j1x j2x þ 2½ j1z , j1y j2y ¼ ½ j1z , j21 j21z þ 2i hj1y j2x 2i hj1x j2y ¼ 2i hð j1y j2x j1x j2y Þ
ð4:42Þ
The commutator is not zero, and so we cannot specify mj1 (or mj2 ) if we specify j. It follows from this analysis that we have to make a choice when specifying the system. Either we use the uncoupled picture jj1mj1 ;j2mj2 i, which leaves the total angular momentum unspecified and therefore, in effect, says nothing about the relative orientation of the two momenta, or we use the coupled picture jj1j2;jmji, which leaves the individual components unspecified. At this stage, which choice we make is arbitrary. Later, when we consider the energy of interaction between different angular momenta we shall see that one picture is more natural than the other. At this stage, the two pictures are simply alternative ways of specifying a composite system.
4.10 The permitted values of the total angular momentum If we decide to use the coupled picture, the question arises as to the permissible values of j and mj. We know that the commutation relations permit j to have any positive integral or half-integral values, but we need to determine which of these many values actually occur for a given j1 and j2. For example, the total angular momentum of a p-electron (l ¼ 1 and s ¼ 12) is unlikely to exceed j ¼ 32. The allowed values of mj follow immediately from the relation jz ¼ j1z þ j2z, and are mj ¼ mj1 þ mj2
ð4:43Þ
114
j
4 ANGULAR MOMENTUM
mj = mj1 + mj2
j2
mj2
j1 mj1
Fig. 4.4 A representation of the requirement that mj ¼ mj1 þ mj2.
That is, the total component of angular momentum about an axis is the sum of the components of the two contributing momenta (Fig. 4.4). To determine the allowed values of j, we first note that the total number of states in the uncoupled picture is (2j1 þ 1)(2j2 þ 1) ¼ 4j1j2 þ 2j1 þ 2j2 þ 1. There is only one state in which both components have their maximum values, mj1 ¼ j1 and mj2 ¼ j2, and this state corresponds to mj ¼ j1 þ j2. However, the maximum value of mj is by definition j, so the maximum value of j is j ¼ j1 þ j2. There are 2j þ 1 ¼ 2j1 þ 2j2 þ 1 states corresponding to this value of j, and so there are a further 4j1j2 states to find. Although the state with mj ¼ j1 þ j2 can arise in only one way, the state with mj ¼ j1 þ j2 1 can arise in two ways, from mj1 ¼ j1 1 and mj2 ¼ j2 and from mj1 ¼ j1 and mj2 ¼ j2 1. The state with j ¼ j1 þ j2 accounts for only one of these states (or for one of their two linear combinations), and so there must be another coupled state for which the maximum value of mj is mj ¼ j1 þ j2 1. This state corresponds to a state with j ¼ j1 þ j2 1. A system with this value of j accounts for a further 2j þ 1 ¼ 2j1 þ 2j2 1 states. The process can be continued by considering the next lower value of mj, which is mj ¼ j1 þ j2 2, and which can be produced in three ways. The two states with j ¼ j1 þ j2 and j ¼ j1 þ j2 1 account for two of them; the third (or the third linear combination) must arise from the state with j ¼ j1 þ j2 2. This argument can be continued, and all the states are accounted for by the time we have reached j ¼ jj1 j2j (j is a positive number, hence the modulus signs). Therefore, the permitted states of angular momentum that can arise from a system composed of two sources of angular momentum are given by the Clebsch–Gordan series: j ¼ j1 þ j2 , j1 þ j2 1, . . ., j j1 j2 j
ð4:44Þ
Example 4.3 Using the Clebsch–Gordan series
What angular momentum states can arise from a system with two sources of angular momentum, one with j1 ¼ 12 and the other with j2 ¼ 32? Specify the states. Method. Use the Clebsch–Gordan series in eqn 4.44 to find the highest and
lowest values of j first, and then complete the series. The composite system has (2j1 þ 1)(2j2 þ 1) states, which may be specified either as jj1mj1 ;j2mj2 i or as jj1j2; jmji. and j12 32j ¼ 1, respectively. So the complete Clebsch–Gordan series is j ¼ 2, 1. A specification of the 4 2 ¼ 8 states in the uncoupled representation is: Answer. The highest and lowest values of j are
1 3 2þ2¼2
j 12 ,þ 12 ; 32 ,þ 32i
j 12 ,þ 12 ; 32 ,þ 12i
j 12 ,þ 12 ; 32 , 12i
j 12 ,þ 12 ; 32 , 32i
j 12 , 12 ; 32 ,þ 32i
j 12 , 12 ; 32 ,þ 12i
j 12 , 12 ; 32 , 12i
j 12 , 12 ; 32 , 32i
The alternative specification, in the coupled representation, is j 12 ,32 ; 2,þ2i j 12 ,32 ; 2,þ1i j 12 ,32 ; 2,0i j 12 ,32 ; 2, 1i j 12 , 32 ; 2, 2i j 12 , 32 ; 1,þ1i j 12 , 32 ; 1,0i
j 12 , 32 ; 1, 1i
4.11 THE VECTOR MODEL OF COUPLED ANGULAR MOMENTA
j
115
Comment. The eight states in the coupled representation are linear combij2
j 1 + j2
nations of the eight states in the uncoupled representation. We explore the relation between them in Section 4.12. j1 + j2 – 1
Self-test 4.3. Repeat the question for j1 ¼ 1 and j2 ¼ 2.
j1 + j 2 – 2
j1
j1 + j2 – 3
|j1 – j2|
Fig. 4.5 The triangle condition corresponding to the Clebsch–Gordan series. The allowed values of j are those for which lines of length j, j1, and j2 can be used to form a triangle. j
The Clebsch–Gordan series can be expressed in a simple pictorial way. Suppose we are given rods of lengths j1 and j2 and are asked for the lengths j of the third side of a triangle that can be formed using these two rods (with all three lengths integers or half-integers). Then the answer would be precisely those given by the Clebsch–Gordan series (Fig. 4.5). For example, j1 ¼ 1 and j2 ¼ 1 require rods of lengths j ¼ 2, 1, 0 to form a triangle. Although the triangle condition is no more than a simple mnemonic, it does suggest that angular momenta in quantum mechanics do in some respects behave like vectors and that the total angular momentum can be regarded as the resultant of the contributing momenta. The exploration of this point leads to the ‘vector model’ of coupled angular momenta.
4.11 The vector model of coupled angular momenta j1
j2
(a)
j
j1 j2
(b)
Fig. 4.6 Two possible states of total angular momentum that can arise from two specified contributing momenta with quantum numbers j1 and j2. The relative orientations of the contributing momenta on their cones determine the total magnitude.
The vector model of coupled angular momentum is an attempt to represent pictorially the features of coupled angular momenta that we have deduced from the commutation relations. The approach gives insight into the significance of various coupling schemes and is often a helpful guide to the imagination: it puts visual flesh on the operator bones. The features that the vector diagrams of coupled momenta must express are as follows: 1. The length of the vector representing the total angular momentum is {j(j þ 1)}1/2, with j one of the values permitted by the Clebsch–Gordan series. 2. This vector must lie at an indeterminate angle on a cone about the z-axis (because jx and jy cannot be specified if jz has been specified). 3. The lengths of the contributing angular momentum vectors are {j1(j1 þ 1)}1/2 and {j2(j2 þ 1)}1/2. These lengths have definite values even when j is specified. 4. The projection of the total angular momentum on the z-axis is mj; in the coupled picture (in which j is specified), the values of mj1 and mj2 are indefinite, but their sum is equal to mj. 5. In the uncoupled picture (in which j is not specified), the individual components mj1 and mj2 may be specified, and their sum is equal to mj. The diagrams in Figs 4.6 and 4.7 capture these points. Figure 4.6 shows one of the states of the uncoupled picture: both mj1 and mj2 are specified, but there is no indication of the relative orientation of j1 and j2 apart from the fact that they lie on their respective cones. The total angular momentum is therefore indeterminate, for it could be either of the resultants shown in (a) or (b) or anything in between. Figure 4.7 shows one of the states of the coupled
116
j
4 ANGULAR MOMENTUM
mj =mj 1 + mj 2 j
mj 1
j1
mj 2 j2
Fig. 4.7 If the two contributing momenta are locked together so that they give rise to a specified total, the projections of the contributing momenta span a range (as depicted by the vertical bars) and although their sum can be specified, their individual values cannot be specified.
picture. Now the resultant, the total angular momentum, has a well-defined magnitude and resultant on the z-axis, but the individual components mj1 and mj2 are indeterminate. It is important not to think of the vectors as actively precessing around their cones: at this stage of describing it, the vector model is a display of possible but unspecifiable orientations. An important example, and one that we shall encounter many times in later chapters, is the case of two particles with spin s ¼ 12, such as two electrons or two protons. For each particle, s ¼ 12 and ms ¼ 12. In the uncoupled picture, the electrons may be in any of the four states a1 a2
a1 b2
b1 a2
b1 b2
These four states are illustrated in Fig. 4.8. The individual angular momenta lie at unspecified positions on their cones and the total angular momentum is indeterminate. Now consider the coupled picture. The triangle condition (or the Clebsch– Gordan series) tells us that the total spin S (upper-case letters are used to denote the angular momenta of collections of particles) can take the values 1 and 0. When S ¼ 0, there is only one possible value of its z-component, namely 0, corresponding to MS ¼ 0. Such a coupled state is called a singlet. When S ¼ 1, MS ¼ þ1, 0, 1, and so this coupled arrangement is called a triplet. The vector model of the triplet is shown in Fig. 4.9. The cones have been drawn to scale, and several points should be apparent. One is that to arrive at a resultant corresponding to S ¼ 1 (of length 21/2) using component vectors corresponding to s ¼ 12 (of length 12 31/2), the vectors must lie at a definite angle relative to one another. In fact, they must lie in the same plane, as shown in the illustration, for only that orientation results in a vector of the correct length. Note that although spins are said to be ‘parallel’ in a triplet state (and represented " "), they are in fact at an acute angle (of close to 70 ). The two spins make the same angle to one another in the states with MS ¼ 1; that is necessary if they are to have the same resultant. The vector model of the singlet must represent a state in which the spin angular momentum vectors sum to give a zero resultant (Fig. 4.10). It is clear from the illustration that the two spins are truly antiparallel ("#) in this state. As in the triplet states, only the relative orientation of the vectors is fixed; the absolute orientation is completely indeterminate.
12
12
12
12
Fig. 4.8 The four uncoupled states of a system consisting of two spin-12 particles (such as electrons), depicted by the cones on which the individual spins lie.
4.12 THE RELATION BETWEEN SCHEMES
j
117
4.12 The relation between schemes The state jj1j2;jmji is built from all values of mj1 and mj2 such that mj1 þ mj2 ¼ mj. This remark suggests that it should be possible to express the coupled state as a sum over all the uncoupled states jj1mj1 ;j2mj2 i that conform to mj1 þ mj2 ¼ mj. It follows that we should be able to write X Cðmj1 , mj2 Þjj1 mj1 ; j2 mj2 i ð4:45Þ jj1 j2 ; jmj i ¼
S = 1, MS = +1
mj1 ;mj2
S = 1, MS = 0
The coefficients C(mj1 ,mj2 ) are called vector coupling coefficients. Alternative names are ‘Clebsch–Gordan coefficients’, ‘Wigner coefficients’, and (in a slightly modified form), the ‘3j-symbols’. We shall illustrate the use of vector coupling coefficients by considering the singlet and triplet states of two spin-12 particles. The values are set out in Table 4.1 (more values for other cases will be found in Appendix 2). The values in the table imply that, using the notation jS,MSi,
S = 1, MS = –1
j1; þ1i ¼ a1 a2 1 1 j1, 0i ¼ 1=2 a1 b2 þ 1=2 b1 a2 2 2 j1, 1i ¼ b1 b2 1 1 j0, 0i ¼ 1=2 a1 b2 1=2 b1 a2 2 2
Fig. 4.9 Three of the four coupled
states of a system consisting of two spin-12 particles. These states all correspond to S ¼ 1. The relative orientations of the individual angular momenta are the same in each case (the angle is arccos (1/3) ¼ 70.53 ).
z
S = 0, MS = 0
Fig. 4.10 The remaining coupled
state of two spin-12 particles. This state corresponds to S ¼ 0. Note that the two contributing momenta are perfectly antiparallel.
There are two points to note. One is that even a ‘spin-parallel’ triplet state (" ") can be composed of ‘opposite’ spins (see the composition of j1,0i). Second, the þ sign in j1,0i is taken to signify that the a and b spins from which it is built are in phase with one another (as suggested by the vector diagram for this state), whereas the sign in j0,0i signifies that they are out of phase. This feature is also captured by the antiparallel arrangement of vectors in the vector diagram. General expressions for the vector coupling coefficients can be derived, but they are very complicated and it is usually simplest to use tables of numerical values. These values can be derived quite simply in special cases, and we shall indicate the procedure for the values in Table 4.1. The general point to note is that the coefficients are in fact the overlap integrals for coupled states with uncoupled states. To see that this is so, multiply both sides Table 4.1 Vector coupling coefficients for s1 ¼ 12, s2 ¼ 12 ms1
ms2
j1, þ1i
j1, 0i
þ12
þ12
1
0
þ12 12 12
12 þ12 12
j0, 0i 0
1/2
0
1/2
0
1/21/2
0
0
j1, 1i 0
1/2
0
1/21/2
0
1/2 0
1
118
j
4 ANGULAR MOMENTUM
of eqn 4.45 from the left by hj1m0j1 ;j2m0j2 j: the only term that survives on the right is the one with mj1 ¼ m0j1 and mj2 ¼ m0j2 (by the orthogonality of the states), so h j1 m0j1 ; j2 m0j2 jj1 j2 ; jmj i ¼ Cðm0j1 , m0j2 Þ
ð4:46Þ
Thus, the coefficient C(mj1 , mj2 ) can be interpreted as the extent to which the coupled state jj1j2; jmji resembles the uncoupled state jj1mj1 ;j2mj2 i. The state j1,þ1i must be composed of a1a2, because only this state corresponds to MS ¼ þ1. It follows that ð4:47Þ
j1, þ1i ¼ a1 a2
The effect of the lowering operator S_ on j1,þ1i is given by eqns 4.23 and 4.29, which in the current notation reads hjS, MS 1i S jS, MS i ¼ fSðS þ 1Þ MS ðMS 1Þg1=2
ð4:48Þ
Therefore hj1, 0i S j1, þ1i ¼ 21=2 However, because S_ ¼ s1 þ s2 , the effect of S_ can also be written hða1 b2 þ b1 a2 Þ S jS, MS i ¼ ðs1 þ s2 Þa1 a2 ¼ Comparison of these two expressions results in j1, 0i ¼
1 ða1 b2 þ b1 a2 Þ 21=2
ð4:49Þ
as found from Table 4.1. The third state of the triplet is obtained by repeating the procedure: hj1, 1i ¼ ðs1 þ s2 Þ S j1, 0i ¼ 21=2
1 ða1 b2 þ b1 a2 Þ 21=2
hb1 b2 ¼ 21=2 It follows that j1, 1i ¼ b1 b2
ð4:50Þ
as we found from the table and exactly as would be expected on physical grounds (namely, that there is only one way of achieving a state with MS ¼ 1 from two spin-12 systems). Only the singlet state remains to be found. Because it necessarily has MS ¼ 0 and MS ¼ ms1 þ ms2 , it is constructed from a1b2 and b1a2. However, it must (by the hermiticity of S2) be orthogonal to the state j1, 0i. Therefore, we can write immediately (to within a factor of 1) that j0, 0i ¼
1 ða1 b2 b1 a2 Þ 21=2
ð4:51Þ
as was given by the use of Table 4.1. As a second illustration, consider two d-electrons. The Clebsch–Gordan series gives the total orbital angular momentum, L, as L ¼ 4, 3, 2, 1, 0. With these states there are associated 25 states, so the problem is somewhat larger than before. The state with L ¼ 4 must have ML ¼ þ4 as one of its
4.13 THE COUPLING OF SEVERAL ANGULAR MOMENTA
j
119
components, and this state can be obtained in only one way, when ml1 ¼ þ2 and ml2 ¼ þ2. It follows that j4,þ4i ¼ j þ2,þ2i where the notation on the left is jL, MLi and that on the right is jml1 , ml2 i. To avoid this rather confusing symbolism, we shall denote the states with L ¼ 0, 1, . . . , 4 by the letters S, P, D, F, G (by analogy with the labels for atomic orbitals). Then instead of the line above we can write jG,þ4i ¼ j þ2,þ2i We may now proceed to generate the remaining eight states with L ¼ 4 by applying the operator L_ ¼ l1 þ l2 . From L_ applied to the left of the last equation we get L jG,þ4i ¼ 81=2 hjG,þ3i and from l1 þ l2 applied to the right we get ðl1 þ l2 Þj þ2,þ2i ¼ 41=2 hðj þ1,þ2i þ j þ2,þ1iÞ from which it follows that 1 jG,þ3i ¼ 1=2 ðj þ1,þ 2i þ j þ2,þ1iÞ 2 The remaining seven states of this set may be generated similarly. The state jF, þ3i also arises from the states j þ1, þ2i and j þ2, þ1i and must be orthogonal to jG, þ 3i. Therefore, we can immediately write 1 jF,þ 3i ¼ 1=2 ðj þ1,þ2i j þ2,þ 1iÞ 2 The remaining six states of this set can now be generated. The same argument may then be applied to generate the D, P, and S states and the table of coefficients given in Appendix 2 can be compiled. Example 4.4 How to use vector coupling coefficients
Construct the state with j ¼ 32 and mj ¼ 12 for a p-electron. Method. For a p-electron, l ¼ 1 and s ¼ 12. The state with j ¼ 32 and mj ¼ 12 is a
linear combination of the states j1,ml;12,msi with ml þ ms ¼ 12. Use Appendix 2 for the vector coupling coefficients. Answer. We write the coupled state in the form
j 32 , 12i ¼
21=2 3
j1, 0; 12 , 12i þ
11=2 3
j1, 1; 12 ,þ 12i
Self-test 4.4. Find the expression for the state jD, 0i arising from the orbital
angular momenta of two p-electrons. Use the tables in Appendix 2.
4.13 The coupling of several angular momenta The final point we need to make in this section concerns the case where three or more momenta are coupled together. In the case of three momenta, we
120
j
4 ANGULAR MOMENTUM
have the choice of first coupling j1 to j2 to form j1,2 and then coupling j3 to that to give the overall resultant j. Illustration 4.2 Coupling several momenta
Consider the total orbital angular momenta of three p-electrons. The coupling of one pair gives l1,2 ¼ 2, 1, 0. Then the third couples with each of these resultants in turn: l1,2 ¼ 2 gives rise to L ¼ 3, 2, 1; l1,2 ¼ 1 gives rise to L ¼ 2, 1, 0; and l1,2 ¼ 0 gives rise to only L ¼ 1. The angular momentum states are therefore F þ 2D þ 3P þ S.
When there are more than two sources of angular momentum, the overall states may be formed in different ways. Thus, instead of the scheme described above, j1 and j3 can first be coupled to form j1,3, and then j2 coupled to j1,3 to form j. The triangle condition applies to each step in the coupling procedure, but the compositions of the states obtained are different. The states obtained by the first coupling procedure can be expressed as linear combinations of the states obtained by the second procedure, and the expansion coefficients are known as Racah coefficients or, in slightly modified form, as ‘6j-symbols’. The question of alternative coupling schemes, and how to select the most appropriate ones, arises in discussions of atomic and molecular spectra, and we shall meet it again there.
PROBLEMS 4.1 Evaluate the commutator [lx,ly] in (a) the position representation, (b) the momentum representation. 4.2 Evaluate the commutators (a) [ly2 ,lx], (b) [ly2 ,lx2 ], and (c) [lx,[lx,ly]]. Hint. Use the basic commutators in eqn 4.7. 4.3 Confirm that [l2, lx] ¼ 0. 4.4 Verify that eqn 4.9 expresses the basic angular momentum commutation rules. Hint. Expand the left of eqn 4.9 and compare coefficients of the unit vectors. Be careful with the ordering of the vector components when expanding the determinant: the operators in the second row always precede those in the third. 4.5 Verify that the five matrices in eqn 4.37 yield the correct results for the applications of the spin operators sq (q ¼ x, y, z, þ , ) on the spin states a and b. 4.6 (a) Confirm that the Pauli matrices 0 1 0 i 1 sx ¼ sy ¼ sz ¼ 1 0 i 0 0
0 1
satisfy the angular momentum commutation relations hsq, and hence provide a matrix when we write sq ¼ 12 representation of angular momentum. (b) Why does the representation correspond to s ¼ 12? Hint. For the second part, form the matrix representing s2 and establish its eigenvalues. 4.7 Using the Pauli matrix representation, reduce each of the operators (a) sxsy, (b) sxs2y s2z , and (c) s2x s2y s2z to a single spin operator. 4.8 Evaluate the effect of (a) eisx =h , (b) eisy =h , (c) eisz =h on an a spin state. Hint. Expand the exponential operators as in Problem 1.15 and use arguments like those in Problem 4.7. 4.9 Suppose that in place of the actual angular momentum commutation rules, the operators obeyed [lx,ly] ¼ i hlz. What would be the roles of l ? 4.10 Calculate the matrix elements (a) h0,0jlzj0,0i, 2 (b) h2,1jlþ j2,0i, (c) h2,2jlþ j2,0i, (d) h2,0jlþ l j2,0i, 2 2 (e) h2,0jl lþ j2,0i, and (f) h2,0jl lz lþ j2,0i.
PROBLEMS
4.11 Demonstrate that j1 j2 is not an angular momentum. 4.12 Calculate the values of the following matrix elements between p-orbitals: (a) hpxjlzjpyi, (b) hpxjlþ jpyi, (c) hpzjlyjpxi, (d) hpzjlxjpyi, and (e) hpzjlxjpxi. 4.13 Evaluate the matrix elements (a) h j; mj þ 1jj3x jj; mj i and (b) h j; mj þ 3jj3x jj; mj i. 4.14 Verify eqn 4.31 for the shift operators in spherical polar coordinates. Use eqn 4.30. 4.15 Confirm that the spherical polar forms of the orbital angular momentum operators in eqn 4.30 satisfy the angular momentum commutation relation [lx, ly] ¼ i hlz and that the shift operators in eqn 4.31 satisfy [lþ , l ] ¼ 2hlz. 4.16 Verify that successive application of l to cll with l ¼ 2 in eqn 4.32 generates the five normalized spherical harmonics Y2ml as set out in Table 3.1. 4.17 (a) Demonstrate that if [j1q, j2q0 ] ¼ 0 for all q, q 0 , then j1 j2 ¼ j2 j1. (b) Go on to show that if j1 j1 ¼ ihj1 and j2 j2 ¼ ihj2, then j j ¼ i hj where j ¼ j1 þ j2. 4.18 In some cases mj1 and mj2 may be specified at the same time as j because although [j2,j1z] is non-zero, the effect of [j2,j1z] on the state with mj1 ¼ j1, mj2 ¼ j2 is zero. Confirm that [j2,j1z]jj1j1; j2j2i ¼ 0 and [j2,j1z]jj1, j1; j2, j2i ¼ 0. 4.19 Determine what total angular momenta may arise in the following composite systems: (a) j1 ¼ 3, j2 ¼ 4; (b) the orbital momenta of two electrons (i) both in p-orbitals, (ii) both in d-orbitals, (iii) the configuration p1d1; (c) the spin angular momenta of four electrons. Hint. Use the Clebsch–Gordan series, eqn 4.44; apply it successively in (c). 4.20 Construct the vector coupling coefficients for a system with j1 ¼ 1 and j2 ¼ 12 and evaluate the matrix elements hj 0 mj 0 jj1zjjmji. Hint. Proceed as in Section 4.12
j
121
and check the answer against the values in Appendix 2. For the matrix element, express the coupled states in the uncoupled representation, and then operate with j1z. 4.21 Use the vector model of angular momentum to derive the value of the angle between the vectors representing (a) two a spins, (b) an a and a b spin in a state with S ¼ 1 and MS ¼ þ1 and MS ¼ 0, respectively. 4.22 Set up a quantum mechanical expression that can be used to derive the same result as in Problem 4.21. Hint. Consider the expectation value of s1 s2. 4.23 Apply both procedures (of the preceding two problems) to calculate the angle between a spins in the aaa state with S ¼ 32. 4.24 Consider a system of two electrons that can have either paired or unpaired spins (e.g. a biradical). The energy of the system depends on the relative orientation of their spins. Show that the operator (hJ/h2)s1 s2 distinguishes between singlet and triplet states. The system is now exposed to a magnetic field in the z-direction. Because the two electrons are in different environments, they experience different local fields and their interaction energy can be written (mB/h)b(g1s1z þ g2s2z) with g1 6¼ g2; mB is the Bohr magneton and g is the electron g-value, quantities discussed in Chapter 13. Establish the matrix of the total hamiltonian, and demonstrate that when hJ >> mBb, the coupled representation is ‘better’, but that when mBb >> hJ, the uncoupled representation is ‘better’. Find the eigenvalues and eigenstates of the system in each case. 4.25 What is the expectation value of the z-component of orbital angular momentum of electron 1 in the jG,MLi state of the configuration d2? Hint. Express the coupled state in terms of the uncoupled states, find hG,MLjl1zjG,MLi in terms of the vector coupling coefficients, and evaluate it for ML ¼ þ4, þ3, . . . , 4. P 4.26 Prove that mj1,mj2 jCmj1,mj2 j2 ¼ 1 for a given j1, j2, j. Hint. Use eqn 4.45 and form hj1j2;jmjjj1j2;jmji.
5 The symmetries of objects 5.1 Symmetry operations and elements 5.2 The classification of molecules The 5.3 5.4 5.5 5.6
calculus of symmetry The definition of a group Group multiplication tables Matrix representations The properties of matrix representations 5.7 The characters of representations 5.8 Characters and classes 5.9 Irreducible representations 5.10 The great and little orthogonality theorems Reduced representations 5.11 The reduction of representations 5.12 Symmetry-adapted bases The symmetry properties of functions 5.13 The transformation of p-orbitals 5.14 The decomposition of directproduct bases 5.15 Direct-product groups 5.16 Vanishing integrals 5.17 Symmetry and degeneracy The full rotation group 5.18 The generators of rotations 5.19 The representation of the full rotation group 5.20 Coupled angular momenta Applications
Group theory
The subject of this chapter—the mathematical theory of symmetry—is one of the most remarkable in quantum mechanics. Not only does it simplify calculations, but it also reveals unexpected connections between apparently disparate phenomena. Whole regions of study are brought together in terms of its concepts. Angular momentum is a part of group theory; so too are the properties of the harmonic oscillator. The conservation of energy and of momentum can be discussed in terms of group theory. Group theory is used to classify the fundamental particles, to discuss the selection rules that govern what spectroscopic transitions are allowed, and to formulate molecular orbitals. The subject simply glitters with power and achievements. What are the capabilities of group theory within quantum chemistry? We shall see that group theory is particularly helpful for deciding whether an integral is zero. Integrals occur throughout quantum chemistry, for they include expectation values, overlap integrals, and matrix elements. It is particularly helpful to know, with minimum effort, whether these integrals are necessarily zero. A limitation of group theory, though, is that it cannot give the magnitude of integrals that it cannot show to be necessarily zero. The values of non-zero integrals typically depend on a variety of fundamental constants, and group theory is silent on them. One particular type of matrix element is the ‘transition dipole moment’ between two states. This quantity determines the intensities of spectroscopic transitions, and if we know that they are necessarily zero, then we have established a selection rule for the transition. In Chapter 2 we encountered the phenomenon of degeneracy and saw qualitatively at least that it is related to the symmetry of the system; group theory lets us anticipate the occurrence and degree of degeneracy that may exist in a system. Finally, we shall see that group theory, by making use of the full symmetry of a system, provides a very powerful way of constructing and classifying molecular orbitals.
The symmetries of objects We begin by establishing the qualitative aspects of the symmetries of objects. This will enable us to classify molecules according to their symmetry. Once molecules have been classified, many properties follow immediately. Moreover, this is a first step to the mathematical formulation of the theory, from which its full power flows.
5.1 SYMMETRY OPERATIONS AND ELEMENTS
j
123
5.1 Symmetry operations and elements An operation applied to an object is an act of doing something to it, such as rotating it through some angle. A symmetry operation is an operation that leaves an object apparently unchanged. For example, the rotation of a sphere around any axis that includes the centre of the sphere leaves it apparently unchanged, and is thus a symmetry operation. The translation of the function sin x through an interval 2p leaves it apparently unchanged, and so it is a symmetry operation of the function. Not all operations are symmetry operations. The rotation of a rectangle through 90 is only a symmetry operation if the rectangle happens to be a square. Every object has at least one symmetry operation: the identity, the operation of doing nothing. To each symmetry operation there corresponds a symmetry element, the point, line, or plane with respect to which the operation is carried out. For example, a rotation is carried out with respect to a line called an ‘axis of symmetry’, and a reflection is carried out with respect to a plane called a ‘mirror plane’. If we disregard translational symmetry operations, then there are five types of symmetry operations that leave the object apparently unchanged, and five corresponding types of symmetry element: E Cn C2,C3,C6
C2
C2 Fig. 5.1 Some of the rotational axes of a regular hexagon, such as a benzene molecule.
The identity operation, the act of doing nothing. The corresponding symmetry element is the object itself. An n-fold rotation, the operation, a rotation by 2p/n around an axis of symmetry, the element.
A hexagon, or a hexagonal molecule such as benzene, has two-, three-, and six-fold axes (C2, C3, and C6, respectively) perpendicular to the plane and several two-fold axes (C2) in the plane (Fig. 5.1). For n > 2 the direction of rotation is significant, and the n orientations of the object are visited in a different order depending on whether the rotation is clockwise as seen from below (Cþ n ) or counterclockwise (Cn ). Therefore, for n > 2, there are two rotations associated with each symmetry axis. If an object (such as a hexagon) has several axes of rotation, then the one with the largest value of n is called the principal axis, provided it is unique. Therefore, for benzene, C6 is the principal axis. s
A reflection, the operation, in a mirror plane, the element.
When the mirror plane includes the principal axis of symmetry, it is termed a vertical plane and denoted sv. If the principal axis is perpendicular to the mirror plane, then the latter symmetry element is called a horizontal plane and denoted sh. A dihedral plane, sd, is a vertical plane that bisects the angle between two C2 axes that lie perpendicular to the principal axis (Fig. 5.2). i
An inversion, the operation, through a centre of symmetry, the element.
124
j
5 GROUP THEORY v h
v'
d C2
(a)
Fig. 5.2 (a) Two vertical mirror
(b)
(c)
C2
planes, (b) a horizontal mirror plane, and (c) a dihedral mirror plane.
The inversion operation is a hypothetical operation which consists of taking each point of an object through its centre and out to an equal distance on the other side (Fig. 5.3).
i
Sn
Fig. 5.3 The centre of inversion of a regular octahedron.
An n-fold improper rotation, the operation (which is also called a ‘rotary-reflection’) occurs about an axis of improper rotation, the symmetry element (or ‘rotary-reflection axis’).
An improper rotation is a composite operation consisting of an n-fold rotation followed by a horizontal reflection in a plane perpendicular to the n-fold axis.1 Neither operation alone is in general a symmetry operation, but the overall outcome is. A methane molecule, for example, has three S4 axes (Fig. 5.4). Care should be taken to recognize improper rotations in disguised form. Thus, S1 is equivalent to a reflection, and S2 is equivalent to an inversion.
5.2 The classification of molecules
S4
S4 Fig. 5.4 An axis of improper rotation in a tetrahedral molecule (such as methane).
To classify a molecule according to its symmetry, we list all its symmetry operations, and then ascribe a label based on the list of those operations. In other words, we use the list of symmetry operations to identify the point group of the molecule. The term ‘point’ indicates that we are considering only the operations corresponding to symmetry elements that intersect in at least one point. That point is not moved by any operation. To classify crystals, we would also need to consider translational symmetry, which would lead us to classify them according to their space group. The name of the point group is expressed using either the Schoenflies system or the International system (which is also called the ‘Hermann– Mauguin system’). It is common to use the former for individual molecules and the latter when considering species in solids. We shall describe and use the Schoenflies system here, but a translation table is given in Table 5.1. In the Schoenflies system, the name of the point group is based on a dominant feature of the symmetry of the molecule, and the label given to the group is in some cases the same as the label of that feature. This double use of a symbol is actually quite helpful, and rarely leads to confusion. .......................................................................................................
1. The order of the operations Cn and sh actually does not matter as these operations commute (Section 5.3).
5.2 THE CLASSIFICATION OF MOLECULES
j
125
Table 5.1 The Schoenflies and International notations for point groups Ci: 1 C1: 1
(a)
T: 23 (b)
Cs : m C2: 2
C3: 3
C4: 4
C6: 6
C2v: 2mm
C3v: 3m
C4v: 4mm
C6v: 6mm
C2h: 2/m
C3h: 6
C4h: 4/m
C6h: 6/m
D2: 222
D3: 32
D4: 422
D6: 622
D2h: mmm
D3h : 62m
D4h: 4/mmm
D6h: 6/mmm
D2d: 42m
D3d: 3m
S4: 4
S6: 3
Td: 43m
Th: m3
O: 432
Oh: m3m
The entries in the table are in the form Schoenflies: International. The International system is also known as the Hermann–Mauguin system. The group D2: 222 is sometimes denoted V and called the Vierer group (group of four).
(c) Fig. 5.5 Objects belonging to the groups (a) C1, (b) Cs, and (c) Ci.
Fig. 5.8 An object belonging to the group C4h.
Fig. 5.6 An object belonging to the group
C4. In this and the following illustrations (up to Fig. 5.15), the shading should not be taken into account when considering
Fig. 5.7 An object belonging to the group C4v.
1. The groups C1, Cs, and Ci. These groups consist of the identity alone (C1), the identity and a reflection (Cs), and the identity and an inversion (Ci) (Fig. 5.5). 2. The groups Cn. These groups consist of the identity and an n-fold rotation (Fig. 5.6). 3. The groups Cnv. In addition to the operations of the groups Cn, these groups also contain n vertical reflections (Fig. 5.7). An important example is the group C1v, the group to which a cone and a heteronuclear diatomic molecule belong. 4. The groups Cnh. In addition to the operations of the groups Cn, these groups contain a horizontal reflection together with whatever operations the presence of these operations imply (Fig. 5.8). It is important to note, as remarked in the last definition, that the presence of a particular set of operations may imply the presence of other operations that are not mentioned explicitly in the definition. For example, C2h automatically possesses an inversion, because rotation by 180 followed by a horizontal reflection is equivalent to an inversion. The full set of operations in each group can be found by referring to the tables (the ‘character tables’) listed in Appendix 1. These tables contain a mass of additional information, and they will gradually move to centre stage as the chapter progresses.
Fig. 5.9 An object belonging to the group D4.
5. The groups Dn. In addition to the operations of the groups Cn, these groups possess n two-fold rotations perpendicular to the n-fold (principal) axis, together with whatever operations the presence of these operations imply (Fig. 5.9).
126
j
5 GROUP THEORY
Fig. 5.10 An object belonging to the
group D4h.
Fig. 5.11 An object belonging to the group D4d.
Fig. 5.12 An object belonging to the group S4.
Fig. 5.13 Objects belonging to the
groups (a) Td, (b) T, and (c) Th.
(a)
(b)
(c)
6. The groups Dnh. These groups consist of the operations present in Dn together with a horizontal reflection, in addition to whatever operations the presence of these operations imply (Fig. 5.10). An important example is D1h, the group to which a uniform cylinder and a homonuclear diatomic molecule belong. 7. The groups Dnd. These groups contain the operations of the groups Dn and n dihedral reflections, together with whatever operations the presence of these operations imply (Fig. 5.11). 8. The groups Sn, with n even. These groups contain the identity and an n-fold improper rotation, together with whatever operations the presence of these operations imply (Fig. 5.12). (a)
(b) Fig. 5.14 Objects belonging to the
groups (a) Oh and (b) O.
Only the even values of n need be considered, because groups with odd n are identical to the groups Cnh, which have already been classified. Note also that the group S2 is equivalent to the group Ci. 9. The cubic and icosahedral groups. These groups contain more than one n-fold rotation with n 3. The cubic groups are labeled T (for tetrahedral) and O for octahedral; the icosahedral group is labelled I. The group Td is the group of the regular tetrahedron; T is the same group but without the reflections of the tetrahedron; Th is a tetrahedral group with an inversion. The group of the regular octahedron is called Oh; if it lacks reflections it is called O. The group of the regular icosahedron is called Ih; if it lacks inversion it is called I. Some objects belonging to these point groups are depicted in Figs 5.13, 5.14, and 5.15, respectively.
5.2 THE CLASSIFICATION OF MOLECULES
j
127
10. The full rotation group, R3. This group consists of all rotations through any angle and in any orientation. It is the symmetry group of the sphere.
Fig. 5.15 An object belonging to the
group I.
Atoms belong to R3, but no molecule does. The properties of R3 turn out to be the properties of angular momentum. This is the deep link between this chapter and Chapter 4, and we explore it later. There are two simple ways of determining to what point group a molecule belongs. One way is to work through the decision tree illustrated in Fig. 5.16. The other is to recognize the group by comparing the molecule with the objects in Fig. 5.17.
Molecule
Y
D∞h
Y
i?
N
Y
C 5?
N
Y
Y
i?
Two or more Cn, n >2?
N
N
Td
Oh
Y
Dnh
N
C∞v
Y
Ih
Linear?
Y
*
Cn? N
N h?
Cs
Y
?
N Dnd
Y
nd?
N
Cnh
N Dn
Y
Ci
Y
the name of a point group to which an object belongs.
S 2n
Y
N
C1
* Select Cn with highest n; then, is nC2 perpendicular to Cn?
nv? N
Fig. 5.16 A flow chart for deciding on
i?
h?
N Cnv
Y
S2n?
N
Cn
128
j
5 GROUP THEORY n =
2
3
4
5
6
∞
Cn
Dn
Cnv (pyramid)
Cone
Cnh
Dnh (plane or bipyramid)
Dnd
S2n
Fig. 5.17 Representative shapes for a variety of point groups.
Example 5.1 How to assign a point group to a molecule
What is the point group of benzene, C6H6? Method. Use the flow chart given in Fig. 5.16, recognizing that benzene has a
unique C6 principal axis that is perpendicular to the molecular plane. Answer. Benzene, a nonlinear molecule, does not contain two (or more) principal axes: C6 is a unique principal axis and there are six C2 axes in the molecular plane and perpendicular to C6; three axes intersect carbon atoms on opposite vertices and three axes bisect carbon–carbon bonds on opposite edges. The molecular plane is sh. From Fig. 5.16, the point group is D6h. Comment. Benzene resembles the hexagon of Fig. 5.17. Self-test 5.1. Assign a point group for 1,4-dichlorobenzene. [D2h]
5.3 THE DEFINITION OF A GROUP
j
129
The calculus of symmetry Power comes to group theory from its mathematical structure. We shall present the material in two stages. The first considers the symmetry operations themselves, and shows how they may be combined together. The second stage shows how to associate matrices with each symmetry operation and to draw on the properties of matrices to establish several important results.
5.3 The definition of a group Symmetry operations can be performed consecutively. We shall use the convention that the operation R followed by the operation S is denoted SR. The order of operations is important because in general the outcome of the operation SR is not the same as the outcome of the operation RS. When the outcomes of RS and SR are equivalent, the operations are said to commute. A general feature of symmetry operations is that the outcome of a joint symmetry operation is always equivalent to a single symmetry operation. We have already seen this property when we saw that a two-fold rotation followed by a reflection in a plane perpendicular to the two-fold axis is equivalent to an inversion: sh C2 ¼ i In general, it is true that for all symmetry operations R and S of an object, we can write RS ¼ T
ð5:1Þ
where T is an operation of the group. A further point about symmetry operations is that there is no difference between the outcomes of the operations (RS)T and R(ST), where (RS) is the outcome of the joint operation S followed by R and (ST) is the outcome of the joint operation T followed by S. In other words, (RS)T ¼ R(ST) and the multiplication of symmetry operations is associative.
Illustration 5.1 Associative property of multiplication of symmetry operations
Consider a square object and the symmetry operations C2 (coincident with the principal C4 axis), i, and sh. Then C2(ish) ¼ C2C2 ¼ E and (C2i)sh ¼ shsh ¼ E and the associative property holds.
These observations, together with two others which are true by inspection, can be summarized as follows: 1. The identity is a symmetry operation. 2. Symmetry operations combine in accord with the associative law of multiplication.
130
j
5 GROUP THEORY
3. If R and S are symmetry operations, then RS is also a symmetry operation. 4. The inverse of each symmetry operation is also a symmetry operation. The third observation implies that R2 (which is shorthand for RR) is a symmetry operation. In observation 4, the inverse of an operation R, generally denoted R1, is defined such that RR1 ¼ R1 R ¼ E
ð5:2Þ
The remarkable point to note is that in mathematics a set of entities called elements form a group if they satisfy the following conditions: v +
C3
v
C 3– v
E
Fig. 5.18 The symmetry elements of
1. 2. 3. 4.
The identity is an element of the set. The elements multiply associatively. If R and S are elements, then RS is also an element of the set. The inverse of each element is a member of the set.
That is, the set of symmetry operations of an object fulfil conditions that ensure they form a group in the mathematical sense. Consequently, the mathematical theory of groups, which is called group theory, may be applied to the study of the symmetry of molecules. This is the justification for the title of this chapter.2
the group C3v.
5.4 Group multiplication tables A table showing the outcome of forming the products RS for all symmetry operations in a group is called a group multiplication table. The procedure used to construct such tables can be illustrated by the group C3v. The symmetry operations for this group are illustrated in Fig. 5.18. We see that there are six members of the group, so it is said to have order 6, which we write as h ¼ 6. To determine the outcome of a sequence of symmetry operations, we consider diagrams like those in Fig. 5.19. You should note that the sequence of changes takes place with respect to fixed positions of the symmetry elements, in the sense that if a Cþ 3 operation is performed, the line representing the sv plane in Fig. 5.18 remains in the same position on the page and is not rotated through 120 by the Cþ 3 operation. Thus it follows that
C 3+
þ þ 00 0 þ C 3 C3 ¼ E sv C3 ¼ sv sv sv ¼ C3
v
v"
The complete set of 36 (in general, h2) products is shown in Table 5.2. As can be seen, each product is equivalent to a single element of the group. Note that RS is not always the same as SR; that is, not all symmetry operations commute. Similar tables can be constructed for all the point groups.
.......................................................................................................
Fig. 5.19 The effect of the operation
Cþ 3 followed by sv is equivalent to the single operation s00v :
2. The unfortunate double meaning of the term ‘element’ should be noted. It is important to distinguish ‘element’, in the sense of a member of a group, from ‘symmetry element’, as defined earlier. The symmetry operations are the elements that comprise the group.
5.5 MATRIX REPRESENTATIONS
j
131
Table 5.2 The C3v group multiplication table E
C3þ
C3
v
s9v
s0v
E
E
Cþ 3
C 3
sv
s0v
s00v
Cþ 3
Cþ 3
C 3
E
s0v
s00v
sv
C 3
C 3
E
s00v
sv
s0v
sv
sv
s00v
E
C 3
Cþ 3
s0v s00v
s0v s00v
sv
Cþ 3 s0v s00v
E
C 3
s0v
v
Cþ 3 C 3
Cþ 3
E
First: Second:
Example 5.2 How to construct a group multiplication table C2
σv
Construct the group multiplication table for the group C2v, the elements of which are shown in Fig. 5.20. 'v
E Fig. 5.20 The symmetry elements of
the group C2v.
Method. Consider a single point on the object of the given point group, and
the effect on the point of each pair of symmetry operations (RS). Identify the single operation that reproduces the effect of the joint application (RS ¼ T), and enter it into the table. Note that ER ¼ RE ¼ R for all R, where E is the identity operation. The orientation on the page of the symmetry elements is unchanged by all the operations. Answer. The group multiplication table is as follows:
E
C2
v
v0
E
C2
sv
sv0
C2
E
sv0
sv
v
sv
sv0
E
C2
v0
sv0
sv
C2
E
E C2
C3
C2
Comment. Note that in this group RS ¼ SR for all entries in the table. Groups C 2"
C 2' E
Fig. 5.21 The symmetry elements of
of this kind, in which the elements commute, are called ‘Abelian’. The group C3v is an example of a ‘non-Abelian group’. Self-test 5.2. Construct the group multiplication table for the group D3, with
elements shown in Fig. 5.21.
the group D3.
5.5 Matrix representations Relations such as RS ¼ T are symbolic summaries of the effect of actions carried out on objects. We can enrich this symbolic representation of symmetry operations by representing the operations by entities that can be manipulated just like ordinary algebra. However, because symmetry
132
j
5 GROUP THEORY
sA
sB
sN sC
Fig. 5.22 One basis for a discussion
of the representation of the group C3v; each sphere can be regarded as an s-orbital centred on an atom.
operations are in general non-commutative (that is, their outcome depends on the order in which they are applied), we should expect to need to use matrices rather than simple numbers, for matrix multiplication is also noncommutative in general. The matrix representative of a symmetry operation is a matrix that reproduces the effect of the symmetry operation (in a manner we describe below). A matrix representation is a set of representatives, one for each element of the group, which multiply together as summarized by the group multiplication table. To establish a matrix representative for a particular operation of a group, we need to choose a basis, a set of functions on which the operation takes place. To illustrate the procedure, we shall consider the set of s-orbitals sA, sB, sC, and sN on an NH3 molecule (Fig. 5.22), which belongs to the group C3v. We have chosen this basis partly because it is simple enough to illustrate a number of points in a straightforward fashion but also because it will be used in the discussion of the electronic structure of an ammonia molecule when we construct molecular orbitals in Chapter 8. The dimension of this basis, the number of members, is 4. We can write the basis as a four-component vector (sN, sA, sB, sC). In general, a basis of dimension d can be written as the row vector f, where f ¼ ðf1 , f2 , . . . , fd Þ Under the operation sv, the vector changes from (sN, sA, sB, sC) to sv(sN, sA, sB, sC) ¼ (sN, sA, sC, sB). This transformation can be represented by a matrix multiplication: 2 3 1 0 0 0 60 1 0 07 7 ð5:3Þ sv ðsN , sA , sB , sC Þ ¼ ðsN , sA , sB , sC Þ6 40 0 0 15 0 0 1 0 This portrayal of the effect of the symmetry operation can be verified by carrying out the matrix multiplication. (For information on matrices, see Further information 23.) The matrix in this expression is the representative of the operation sv for the chosen basis, and is denoted D(sv). Note that a fourdimensional basis gives rise to a 4 4-dimensional representative, and that in general a d-dimensional basis gives rise to a d d-dimensional representative. In terms of the explicit rules for matrix multiplication, the effect of an operation R on the general basis f is to convert the component fi into X fj Dji ðRÞ ð5:4Þ Rfi ¼ j
where Dji(R) is a matrix element of the representative D(R) of the operation R. For example, sv sB ¼ sN 0 þ sA 0 þ sB 0 þ sC 1 ¼ sC as required. The representatives of the other operations of the group can be found in the same way. Note that because Ef ¼ f, the representative of the identity operation is always the unit matrix.
5.5 MATRIX REPRESENTATIONS
j
133
Example 5.3 How to formulate a matrix representative
Find the matrix representative for the operation Cþ 3 in the group C3v for the s-orbital basis used above. Method. Examine Fig. 5.22 to decide how each member of the basis is
transformed under the operation, and write this transformation in the form Rf ¼ f 0 . Then construct a d d matrix D(R) which generates f 0 when fD(R) is formed and multiplied out. Answer. Inspection of Fig. 5.22 shows that under the operation,
Cþ 3 ðsN , sA , sB , sC Þ ¼ ðsN , sB , sC , sA Þ This transformation can be expressed as 2 1 60 þ 6 C3 ðsN , sA , sB , sC Þ ¼ ðsN , sA , sB , sC Þ4 0 0
the matrix product 3 0 0 0 0 0 17 7 1 0 05 0 1 0
Therefore, the 4 4 matrix above is the representative of the operation Cþ 3 in the basis. Self-test 5.3. Find the matrix representative of the operation C 3 in the same
basis. [Table 5.3]
The complete set of representatives for this basis are displayed in Table 5.3. We now arrive at a centrally important point. Consider the effect of the consecutive operations Cþ 3 followed by sv. From the group multiplication table we know that the effect of the joint operation svCþ 3 is the same as the effect of the reflection sv00 . That is, 00 sv Cþ 3 ¼ sv
Table 5.3 The matrix representation of C3v in the basis {sN,sA,sB,sC} D(E) 2 1 0 60 1 6 40 0 0 0
0 0 1 0
3
0 07 7 05 1
w(sv) ¼ 2
0 0 0 1
3
0 17 7 05 0
(Cþ 3 )¼1
w(E) ¼ 4 D(s ) 2 v 1 0 60 1 6 40 0 0 0
D (Cþ 3) 2 1 0 60 0 6 40 1 0 0
0 0 0 1
3 0 07 7 15 0
D(s 0 ) 2 v 1 0 60 0 6 40 1 0 0 (sv0 ) ¼ 2
0 1 0 0
D(C ) 2 3 1 0 60 0 6 40 0 0 1
0 1 0 0
3 0 07 7 15 0
(C 3)¼1 3 0 07 7 05 1
D(s00 ) 2 v 1 0 60 0 6 40 0 0 1
0 0 1 0
(sv00 ) ¼ 2
3 0 17 7 05 0
134
j
5 GROUP THEORY
Now consider this joint operation in terms of 2 32 1 0 0 0 1 0 0 6 0 1 0 0 76 0 0 0 6 76 Dðsv ÞDðCþ 76 3Þ ¼ 6 4 0 0 0 1 54 0 1 0 0 0 1 0 0 0 1
the matrix representatives. 3 2 3 0 1 0 0 0 6 7 17 7 60 0 0 17 7¼6 7 05 40 0 1 05 0
0 1 0 0
¼ Dðs00v Þ That is, the matrix representatives multiply together in exactly the same way as the operations of the group. This is true whichever operations are considered, and so the set of six 4 4 matrices in Table 5.3 form a matrix representation of the group for the selected basis in the sense that if RS ¼ T, then DðRÞDðSÞ ¼ DðTÞ
ð5:5Þ
for all members of the group. Proof 5.1 The representation of group multiplication
The formal proof that the representatives multiply in the same way as the symmetry operations gives a taste of the kind of manipulation that will be needed later. Once again we consider two elements R and S which multiply together to give the element T. It follows from eqn 5.4 that for the general basis f, X X fj Dji ðSÞ ¼ fk Dkj ðRÞDji ðSÞ RSfi ¼ R j j, k The sum over j of Dkj(R) Dji(S) is the definition of a matrix product, and so X RSfi ¼ fk fDðRÞDðSÞgki k
where {D(R)D(S)}ki refers to the element in row k and column i of the matrix given by the product D(R)D(S). However, we also know that RS ¼ T, so we can also write X RSfi ¼ Tfi ¼ fk fDðTÞgki k
By comparing the two equations we see that fDðRÞDðSÞgki ¼ fDðTÞgki for all elements k and i. Therefore, DðRÞDðSÞ ¼ DðTÞ That is, the representatives do indeed multiply like the group elements, as we set out to prove.
It follows from the fact that the representatives multiply like the group elements, that the representatives of an operation R and its inverse R1 are related by DðR1 Þ ¼ DðRÞ1 1
ð5:6Þ
where D denotes the inverse of the matrix D. For instance, because RR1 ¼ E, it follows that DðRÞDðR1 Þ ¼ DðRÞDðRÞ1 ¼ 1 ¼ DðEÞ where 1 is the unit matrix.
5.6 THE PROPERTIES OF MATRIX REPRESENTATIONS
j
135
5.6 The properties of matrix representations
s1
s2
To develop the content of matrix representations, we need to introduce some of their properties. In each case we shall introduce the concept using the s-orbital basis for C3v to fix our ideas, and then generalize the concept to any basis for any group. To begin, we introduce the concept of ‘similarity transformation’. Suppose that instead of the s-orbital basis, we select a linear combination of these orbitals to serve as the basis. One such set might be (sN, s1, s2, s3), where s1 ¼ sA þ sB þ sC, s2 ¼ 2sA sB sC, and s3 ¼ sB sC (apart from the requirement that the combinations are linearly independent, the choice is arbitrary, but later we shall see that this set has a special significance). The combinations are illustrated in Fig. 5.23. We should expect the matrix representation in this basis to be similar to that in the original basis. This similarity is given a formal definition by saying that two representations are similar if the representatives for the two bases are related by the similarity transformation DðRÞ ¼ cD0 ðRÞc1
s3
ð5:7aÞ
where c is the matrix formed by the coefficients relating the two bases (see the proof below for an explicit definition). The inverse relation is obtained by multiplication from the left by c1 and from the right by c: D0 ðRÞ ¼ c1 DðRÞc
Fig. 5.23 The symmetry-adapted
linear combinations of the peripheral atom orbitals in a C3v molecule.
ð5:7bÞ
Proof 5.2 The similarity of representations
Because the new basis f 0 ¼ (f10 , f20 , . . . , fd0 ) is a linear combination of the original basis f ¼ (f1, f2, . . . , fd), we can express any member as X fj cji fi0 ¼ j
where the cji are constant coefficients.3 This expansion can be expressed as a matrix product by writing f 0 ¼ fc where c is the matrix formed of the coefficients cji. Now suppose that in the original basis the representative of the element R is D(R) in the sense that X Rfi ¼ fk Dki ðRÞ, or Rf ¼ fDðRÞ k
Likewise, the effect of the same operation on a member of the transformed basis set is Rfi0 ¼
X
fk0 D0ki ðRÞ,
or Rf 0 ¼ f 0 D0 ðRÞ
k
.......................................................................................................
3. For the particular basis f 0 ¼ (sN, s1, s2, s3), the coefficients are specified in Example 5.4.
136
j
5 GROUP THEORY
In general, if two matrices A and B are related by an expression of the form A ¼ CBC1, then the matrices are said to be similar and the expression is a similarity transformation. Such transformations are useful in diagonalizing matrices as encountered in Example 1.10.
The relation between the two ‘similar’ representatives can be found by substituting f 0 ¼ fc into the last equation, which then becomes Rfc ¼ fcD0 ðRÞ If we then multiply through from the right by c1, the reciprocal of the matrix c (in the sense that cc1 ¼ c1c ¼ 1), then we obtain Rf ¼ fcD0 ðRÞc1 Comparison of this expression with Rf ¼ fD(R) leads to eqn 5.7a.
Example 5.4 How to construct a similarity transformation
The representative of the operation Cþ 3 in C3v for the s-orbital basis is given in Table 5.3. Derive an expression for the representative in the transformed basis given at the start of this subsection. Method. To implement the recipe in eqn 5.7, we need to construct the matrices
c and c1. Therefore, begin by expressing the relation between the two bases in matrix form (as f 0 ¼ fc), and find the reciprocal of c by the methods described in Further information 23. Finally, evaluate the matrix product c1D(R)c. Answer. The relation between the two bases,
sN ¼ sN
s1 ¼ sA þ sB þ sC
s2 ¼ 2sA sB sC
s3 ¼ sB sC
can be expressed as the following matrix: 2 3 1 0 0 0 60 1 2 0 7 7 ðsN , s1 , s2 , s3 Þ ¼ ðsN , sA , sB , sC Þ6 4 0 1 1 1 5 0 1 1 1 which lets us identify the matrix c. The reciprocal of this matrix is 2 3 6 0 0 0 60 2 2 2 7 7 c1 ¼ 16 6 4 0 2 1 1 5 0 0 3 3 The representative of Cþ 3 in the new basis is therefore 1 þ D0ðCþ 3 Þ ¼ c 2 DðC3 Þc 32 6 0 0 0 1 0 60 2 2 76 0 0 2 6 76 ¼ 16 6 76 4 0 2 1 1 54 0 1 0 0 3 3 0 0 2 3 2 1 6 0 0 0 60 6 0 7 6 0 7 60 6 ¼ 16 6 7¼6 4 0 0 3 3 5 4 0
0 0
9
3
0
0 0 0 1
32 0 1 60 17 76 76 0 54 0 0
0
0 1
0 2
1 1 1 1 3
3 0 0 7 7 7 1 5 1
0 0 0 1 0 0 7 7 7 1 0 2 12 5 0
3 2
12
Self-test 5.4. Find the representative for the operation sv in the transformed
basis. [See Table 5.4]
5.7 THE CHARACTERS OF REPRESENTATIONS
j
137
Table 5.4 The matrix representation of C3v in the basis {sN,s1,s2,s3} D(E) 2 1 0 60 1 6 40 0 0 0
D(Cþ 3) 2 1 0 60 1 6 6 40 0 0 0
3 0 07 7 05 1
0 0 1 0
w(E) ¼ 4 D(sv) 2 1 0 0 60 1 0 6 40 0 1 0 0 0 w(sv) ¼ 2
3
0 0 7 7 0 5 1
3
0 0 0 0 7 7 7 12 12 5 1 12 2
D(C 3) 3 2 1 0 0 0 60 1 0 0 7 7 6 7 6 4 0 0 12 12 5 0 0 12 12
(Cþ 3 )¼1
(C 3 )¼1
D(s0v ) 3 2 1 0 0 0 60 1 0 07 7 6 7 6 4 0 0 12 12 5 3 1 0 0 2 2 (s0v ) ¼ 2
D(s00 ) 3 2 v 1 0 0 0 60 1 0 0 7 7 6 7 6 1 4 0 0 2 12 5 3 1 0 0 2 2 (s00v ) ¼ 2
The same technique as that illustrated in the example may be applied to the other representatives, and the results are collected in Table 5.4.
5.7 The characters of representations There is one striking feature of the two representations in Tables 5.3 and 5.4. Although the matrices differ for the two bases, for a given operation the sum of the diagonal elements of the representative is the same in the two bases. The diagonal sum of matrix elements is called the character of the matrix, and is denoted by the symbol w(R) where w is chi: X Dii ðRÞ ð5:8Þ wðRÞ ¼ i
In matrix algebra, the sum of diagonal elements is called the trace of the matrix, and denoted tr. So, a succinct definition of the character of the operation R is wðRÞ ¼ tr DðRÞ ð5:9Þ We now demonstrate that the character of an operation is invariant under a similarity transformation of the basis. The proof makes use of the fact (which we shall use several times in the following discussion) that the trace of a product of matrices is invariant under cyclic permutation of the matrices: tr ABC ¼ tr CAB ¼ tr BCA ð5:10Þ Proof 5.3 The invariance of the trace of a matrix and the character of a representative
First, we express the trace as a diagonal sum: X tr ABC ¼ ðABCÞii i
Then we expand the matrix product by the rules of matrix multiplication: X Aij Bjk Cki tr ABC ¼ ijk
138
j
5 GROUP THEORY
Matrix elements are simple numbers that may be multiplied in any order. If they are permuted cyclically in this expression, neighbouring subscripts continue to match, and so the matrix product may be reformulated with the matrices in a permuted order: X X tr ABC ¼ Bjk Cki Aij ¼ ðBCAÞjj ¼ tr BCA ijk
j
as required. Now we apply this general result to establish the invariance of the character under a similarity transformation brought about by the matrix c: wðRÞ ¼ tr DðRÞ ¼ tr cD0 ðRÞc1 ¼ tr D0 ðRÞc1 c ¼ tr D0 ðRÞ ¼ w0 ðRÞ That is, the characters of R in the two representations, w(R) and w 0 (R), are equal, as we wanted to prove.
5.8 Characters and classes One feature of the characters shown in Tables 5.3 and 5.4 is that the characters of the two rotations are the same, as are the characters of the three reflections. These equalities suggest that the operations fall into various classes that can be distinguished by their characters. The formal definition of the class of a symmetry operation is that two operations R and R 0 belong to the same class if there is some symmetry operation S of the group such that R0 ¼ S1 RS
ð5:11Þ
The elements R and R 0 are said to be conjugate. Conjugate members belong to the same class. The physical interpretation of conjugacy and membership within a class is that R and R 0 are the same kind of operation (such as a rotation) but performed with respect to symmetry elements that are related by a symmetry operation.
Example 5.5 How to show that two symmetry operations are conjugate Show that the symmetry operations Cþ 3 and C3 are conjugate in the group C3v.
Method. We need to show that there is a symmetry transformation of the group that transforms Cþ 3 into C3 . Intuitively, we know that the reflection of a rotation in a vertical plane reverses the sense of the rotation, so we can suspect that a reflection is the necessary operation. To work out the effect of a succession of operations, we use the information in the group multiplication table (Table 5.2); to find the reciprocal of an operation, we look for the element that produces the identity E in the group multiplication table.
5.9 IRREDUCIBLE REPRESENTATIONS
j
139
þ Answer. We consider the joint operation s1 v C3 sv. According to Table 5.2,
the inverse of sv is sv itself. Therefore, from the group multiplication table we can write þ þ 0 s1 v C3 sv ¼ sv ðC3 sv Þ ¼ sv sv ¼ C3
Hence, the two rotations belong to the same class. Self-test 5.5. Show that sv and sv0 are members of the same class in C3v.
With the concept of conjugacy established, it is now straightforward to demonstrate that symmetry operations in the same class have the same character in a given representation. Proof 5.4 The invariance of character
We use the cyclic invariance of the trace of the product of representatives (eqn 5.10). We also use the fact (as a result of eqn 5.11) that D(R 0 ) and D(R) are related by a similarity transformation: wðR0 Þ ¼ tr DðR0 Þ ¼ tr D1 ðSÞDðRÞDðSÞ ¼ tr DðRÞDðSÞD1 ðSÞ ¼ tr DðRÞ ¼ wðRÞ
A word of warning: although it is true that all members of the same class have the same character in a given representation, the characters of different classes may be the same as one another. For example, as we shall see, one matrix representation of a group consists of 1 1 matrices each with the single element 1. Such a representation certainly reproduces the group multiplication table, but does so in a trivial way, and hence is called the unfaithful representation of the group. We shall see later that this representation is in fact one of the most important of all possible representations. The characters of all the operations of the group are 1 in the unfaithful representation, and although it is true that members of the same class have the same character (1 in each case), different classes also share that character.
5.9 Irreducible representations Inspection of the representation of the group C3v in Table 5.3 for the original s-orbital basis shows that all the matrices have a block-diagonal form: 2 3 1 0 0 0 60 7 6 7 40 5 0 As a consequence, we see that the original four-dimensional basis may be broken into two, one consisting of sN alone and the other of the
140
j
5 GROUP THEORY
three-dimensional basis (sA, sB, sC): E 1
2
1 6 40 0
Cþ 3 1
3 2 3 0 0 0 0 1 7 6 7 1 05 41 0 05 0 1 0 1 0
s0v sv 1 1 2 3 2 3 1 0 0 0 1 0 40 0 15 41 0 05 0 1 0 0 0 1
C 3 1
2
3 0 1 0 6 7 40 0 15 1 0 0 s00v 1 2 3 0 0 1 40 1 05 1 0 0
The first row in each case is the one-dimensional representation spanned by sN and the 3 3 matrices form the three-dimensional representation spanned by the three-dimensional basis (sA, sB, sC). The separation of the representation into sets of matrices of lower dimension is called the reduction of the representation. In this case, we write Dð4Þ ¼ Dð3Þ Dð1Þ
ð5:12Þ
and say that the four-dimensional representation has been reduced to a direct sum (the significance of the sign) of a three-dimensional and a onedimensional representation. The term ‘direct sum’ is used because we are not simply adding together matrices in the normal way but creating a matrix of high dimension from matrices of lower dimension. There are several points that should be noted about the reduction. First, we see that one of the representations obtained is the unfaithful representation mentioned earlier, in which all the representatives are 1 1 matrices with the same single element, 1, in each case. Another point is that the characters of the representatives of symmetry operations of the same class are the same, as we proved earlier. That is true of D(4), D(3), and D(1) (although the characters do have different values for each representation). The question that we now confront is whether D(3) is itself reducible. A glance at the representation in Table 5.4 shows that the similarity transformation we discussed earlier converts D(4) to a block-diagonal form of structure 2 3 1 0 0 0 60 1 0 07 6 7 40 0 5 0 0 which corresponds to the reduction Dð4Þ ¼ Dð1Þ Dð1Þ Dð2Þ The two one-dimensional representations in this expression are the same as the single one-dimensional (and unfaithful) representation introduced above, so in effect the new feature we have achieved is the reduction of the
5.9 IRREDUCIBLE REPRESENTATIONS
j
141
three-dimensional representation: Dð3Þ ¼ Dð1Þ Dð2Þ In this case, the linear combination s1 is a basis for D(1) whereas before the single orbital sN was a basis for D(1). A glance at Fig. 5.23 shows the physical reason for this analogy: the orbital sN has the ‘same symmetry’ as s1. However, we are now moving to a position where we can say what we mean by the colloquial term ‘same symmetry’: we mean act as a basis of the same matrix representation. The question that immediately arises is whether the two-dimensional representation can be reduced to the direct sum of two one-dimensional representations by another choice of similarity transformation. As we shall see shortly, group theory can be used to confirm that D(2) is an irreducible representation (‘irrep’) of the molecular point group in the sense that no similarity transformation (that is, linear combination of basis functions) can be found that simultaneously converts the representatives to block-diagonal form. The unfaithful representation D(1) is another example of an irreducible representation. Each irreducible representation of a group has a label called a symmetry species. The symmetry species is ascribed on the basis of the list of characters of the representation. Thus, the unfaithful representation of the group C3v has the list of characters (1, 1, 1, 1, 1, 1) and belongs to the symmetry species named A1.4 The two-dimensional irreducible representation has characters (2, 1, 1, 0, 0, 0), and its label is E. The letters A and B are used for the symmetry species of one-dimensional irreducible representations, E is used for two-dimensional irreducible representations, and T is used for threedimensional irreducible representations. The irreducible representations labelled A1 and E are also labelled G(1) and G(3), respectively (we meet G(2) shortly: the numbers on G do not refer to the dimension of the irreducible representation, they are just labels). We shall use the G notation for general expressions and the A, B, . . . labels in particular cases. If a particular set of functions is a basis for an irreducible representation G, then we say that the basis spans that irreducible representation. The complete list of characters of all possible irreducible representations of a group is called a character table. As we shall shortly show, there are only a finite number of irreducible representations for groups of finite order, and we shall see that these tables are of enormous importance and usefulness. We are now left with three tasks. One is to determine which symmetry species of irreducible representation may occur in a group and establish their characters. The second is to determine to what direct sum of irreducible representations an arbitrary matrix representation can be reduced—that is equivalent to deciding which irreducible representations an arbitrary basis spans. The third is to construct the linear combinations of members of an arbitrary basis that span a particular irreducible representation. This work requires some powerful machinery, which the next subsection provides. .......................................................................................................
4. For any point group, the unfaithful representation will be labelled with the letter A.
142
j
5 GROUP THEORY
5.10 The great and little orthogonality theorems The quantitative development of group theory is based on the great orthogonality theorem (GOT), which states the following. Consider a group of order h, and let D(l)(R) be the representative of the operation R in a dl-dimensional irreducible representation of symmetry species G(l) of the group. Then X ðlÞ h ðl0 Þ Dij ðRÞ Di0 j0 ðRÞ ¼ dll0 dii0 djj0 ð5:13Þ d l R Note that this form of the theorem allows for the possibility that the representatives have complex elements; in the applications in this chapter, however, they will in fact be real and complex conjugation has no effect. Although this expression may look fearsome, it is simple to apply. In words, it states that if you select any location in a matrix of one irreducible representation, and any location in a matrix of the same or different irreducible representation of the group, multiply together the numbers found in those two locations, and then sum the products over all the operations of the group, then the answer is zero unless the locations of the elements are the same in both sets of matrices, and indeed the same set of matrices (the same irreducible representations) are chosen. If the locations are the same, and the two irreducible representations are the same, then the result of the calculation is h/dl. Example 5.6 How to use the great orthogonality theorem
Illustrate the validity of the GOT by choosing two examples from Table 5.3, one that gives a non-zero value and one that gives a zero value according to the theorem. Method. For a non-zero outcome, we must choose the same location
in the same matrix representation: a simple example would be to use the onedimensional unfaithful representation A1. For the zero outcome, we can choose either different locations in a single irreducible representation or arbitrary locations in two different irreducible representations. Refer to Table 5.3 for the specific values of the matrix elements. Answer. (a) For C3v, for which h ¼ 6, take the irreducible representation A1 (which has d ¼ 1), in which the matrices are 1, 1, 1, 1, 1, 1. The sum on the left of the GOT with each matrix element multiplied by itself is X ðA Þ ðA Þ D111 ðRÞ D111 ðRÞ ¼ 1 1 þ 1 1 þ 1 1 þ 1 1 þ 1 1 þ 1 1 ¼ 6 R
which is equal to 6/1 ¼ 6, as required by the theorem. (b) Consider two different locations in the two-dimensional irreducible representation E. For example, take the 34 and 33 elements of the matrices in Table 5.3: X ðEÞ ðEÞ ðEÞ ðEÞ ðEÞ ðEÞ þ D34 ðRÞ D33 ðRÞ ¼ D34 ðEÞ D33 ðEÞ þ D34 ðCþ 3 Þ D33 ðC3 Þ þ R
¼0 1þ0 0þ1 0þ1 0þ0 0þ0 1¼0 which is also in accord with the theorem.
5.10 THE GREAT AND LITTLE ORTHOGONALITY THEOREMS
j
143
Self-test 5.6. Confirm the validity of the GOT by using the irreducible
representation A1 and any element of the irreducible representation E for the matrices in Table 5.4.
The great orthogonality theorem is too great for most of our purposes, and it is possible to derive from it a weaker statement in terms of the characters of irreducible representations. The little orthogonality theorem (LOT) states that X 0 wðlÞ ðRÞ wðl Þ ðRÞ ¼ hdll0 ð5:14Þ R
Proof 5.5 The little orthogonality theorem
To prove the little orthogonality theorem from the GOT, we set j ¼ i and j 0 ¼ i 0 , to obtain diagonal elements on the left of eqn 5.13, and then sum over all these diagonal elements. The left of eqn 5.13 becomes ( )( ) X X ðlÞ X ðl0 Þ X X ðlÞ ðl0 Þ Dii ðRÞ Di0 i0 ðRÞ ¼ Dii ðRÞ Di0 i0 ðRÞ i;i0
R
¼
i0
i
R
X
ðlÞ
ðl0 Þ
w ðRÞ w ðRÞ
R
Under the same manipulations, the right-hand side of eqn 5.13 becomes X X h h dll0 dii0 dii0 ¼ dll0 dii d d l l i;i0 i There are dl values of the index i in a matrix of dimension dl, and so the sum on the right is the sum of 1 taken dl times, or dl itself. Hence, on combining the two halves of the equation, we arrive at the little orthogonality theorem.
The LOT can be expressed slightly more simply by making use of the fact that all operations of the same class have the same character. Suppose that the number of symmetry operations in a class c is g(c), so that g(C3) ¼ 2 and g(sv) ¼ 3 in the group C3v. Then X 0 gðcÞwðlÞ ðcÞ wðl Þ ðcÞ ¼ hdll0 ð5:15Þ c
where the sum is now over the classes. When l 0 ¼ l, this expression becomes X 2 gðcÞ wðlÞ ðcÞ ¼ h ð5:16Þ c
which signifies that the sum of the squares of the characters of any irreducible representation of a group is equal to the order of the group. The form of the LOT suggests the following analogy. Suppose we interpret the quantity {g(c)}1/2wc(l) as a component vc(l) of a vector v(l), with
144
j
5 GROUP THEORY
each component distinguished by the index c; then the LOT can be written X 0 ðl0 Þ ðlÞ vðlÞ vðl Þ ¼ hdll0 ð5:17Þ c vc ¼ v c
This expression shows that the LOT is equivalent to the statement that two vectors are orthogonal unless l 0 ¼ l. However, the number of orthogonal vectors in a space of dimension N cannot exceed N (think of the three orthogonal vectors in ordinary space). In the present case, the dimensionality of the ‘space’ occupied by the vectors is equal to the number of classes of the group. Therefore, the number of values of l which distinguish the different orthogonal vectors cannot exceed the number of classes of the group. Because l labels the symmetry species of the irreducible representations of the group, it follows that the number of symmetry species cannot exceed the number of classes of the group. In fact, it follows from a more detailed analysis of the GOT (as distinct from the LOT) that these two numbers are equal. Hence, we arrive at the following restriction on the structure of a group: The number of symmetry species is equal to the number of classes. The vector interpretation can be applied to the GOT itself. To do so, we identify Dij(l)(R) as the Rth component of a vector v identified by the three indices l, i, and j. The orthogonality condition is then 0 0 0
vðl;i;jÞ vðl ;i ;j Þ ¼
h dll0 dii0 djj0 dl
ð5:18Þ
This condition implies that any pair of vectors with different labels are orthogonal. The orthogonality condition is expressed in terms of a sum over all h elements of a group, so the vectors are h-dimensional. The total number of vectors of a given irreducible representation is dl2 because the labels i and j can each take dl values in a dl dl matrix. The total dimensionality of the 2 space is therefore
P 2 the sum of dl over all the symmetry species. The resulting cannot exceed the dimension h of the space the vectors number l dl inhabit, and it may be shown that the two numbers are in fact equal. Therefore, we have the following further restriction on the structure of the group: X dl2 ¼ h ð5:19Þ l
Example 5.7 How to construct a character table
Use the restrictions derived above and the LOT to complete the C3v character table. Method. We have identified two of the irreducible representations of the six-
dimensional group, namely A1 and E. The restriction given above will tell us the number of symmetry species to look for, and we can use eqn 5.19 to determine their dimensions. The characters themselves can be found from the LOT by ensuring that they are orthogonal to the two irreducible representations we have already found.
5.10 THE GREAT AND LITTLE ORTHOGONALITY THEOREMS
j
145
Answer. The order of the group is h ¼ 6 and there are three classes of operation; therefore, we expect there to be three symmetry species of irreducible representation. The dimensionality, d, of the unidentified irreducible representation must satisfy
12 þ 22 þ d2 ¼ 6 Hence, d ¼ 1, and the missing irreducible representation is one-dimensional. We shall call it A2. At this stage we can use the LOT to construct three equations for the three unknown characters. With l ¼ l 0 ¼ A2, eqn 5.16 is fwðA2 Þ ðEÞg2 þ 2fwðA2 Þ ðC3 Þg2 þ 3fwðA2 Þ ðsv Þg2 ¼ 6 With l ¼ A2 and l 0 ¼ A1 we obtain
Table 5.5 The C3v character
table
wðA2 Þ ðEÞwðA1 Þ ðEÞ þ 2wðA2 Þ ðC3 ÞwðA1 Þ ðC3 Þ þ 3wðA2 Þ ðsv ÞwðA1 Þ ðsv Þ ¼ 0
C3v
E
2C3
3 v
A1
1
1
1
A2
1
1
1
E
2
1
0
and with l ¼ A2 and l 0 ¼ E wðA2 Þ ðEÞwðEÞ ðEÞ þ 2wðA2 Þ ðC3 ÞwðEÞ ðC3 Þ þ 3wðA2 Þ ðsv ÞwðEÞ ðsv Þ ¼ 0 When the known values of the characters of A1 and E are substituted, these two equations become wðA2 Þ ðEÞ þ 2wðA2 Þ ðC3 Þ þ 3wðA2 Þ ðsv Þ ¼ 0 and 2wðA2 Þ ðEÞ 2wðA2 Þ ðC3 Þ ¼ 0
Table 5.6 The C2v character
table 0v
C2v
E
C2
A1
1
1
1
1
A2
1
1
1
1
B1
1
1
1
1
B2
1
1
1
1
v
The three equations are enough to determine the three unknown characters, and we find wðA2 Þ (E) ¼ 1, wðA2 Þ (C3) ¼ 1, and wðA2 Þ (sv) ¼ 1. The complete set of characters is displayed in Table 5.5. Comment. The character of the identity in a one-dimensional irreducible
representation is 1, so that value could have been obtained without any calculation. Self-test 5.7. Construct the character table for the group C2v. [See Table 5.6]
The character table for any symmetry group can be constructed as we have illustrated, and a selection of character tables is given in Appendix 1.
Reduced representations A great deal depends on being able to establish what irreducible representations are spanned by a given basis. This problem leads us into the applications of group theory that we shall use throughout the text.
146
j
5 GROUP THEORY
5.11 The reduction of representations The question we now tackle is, given a general set of basis functions, how do we find the symmetry species of the irreducible representations they span? Often, as we shall see, we are interested more in the symmetry species and its characters than in the actual irreducible representation (the set of matrices). We have seen that a representation may be expressed as a direct sum of irreducible representations DðRÞ ¼ DðG D
ð1Þ
Þ
ðRÞ DðG
ð2Þ
Þ
ðRÞ
ð5:20Þ
by finding a similarity transformation that simultaneously converts the matrix representatives to block-diagonal form. It is notationally simpler to express this reduction in terms of the symmetry species of the irreducible representations that occur in the reduction: X G¼ al GðlÞ ð5:21Þ
D (1)
l
D (2) D (3) (1) + (2) + (3) Fig. 5.24 A diagrammatic
representation of the reduction of a matrix to block-diagonal form. The sum of the diagonal elements remains unchanged by the reduction.
where al is the number of times the irreducible representation of symmetry species G(l) appears in the direct sum. For example, the reduction of the s-orbital basis we have been considering would be written G ¼ 2A1 þ E. Our task is to find the coefficients al. To do so, we make use of the fact that because the character of an operation is invariant under a similarity transformation, the character of the original representative is the sum of the characters of the irreducible representations into which it is reduced (Fig. 5.24). Therefore, X al wðlÞ ðRÞ ð5:22Þ wðRÞ ¼ l
Now we use the LOT to determine the coefficients. To do so, we multiply both 0 sides of this equation by w(l )(R) and sum over all the elements of the group: X X 0 X 0 wðl Þ ðRÞ wðRÞ ¼ al wðl Þ ðRÞ wðlÞ ðRÞ R
R
¼h
X
l
al dll0 ¼ hal0
l
That is, the coefficients are given by the rule 1 X ðlÞ w ðRÞ wðRÞ al ¼ h R
ð5:23Þ
Because the characters of members of the same class of operation are the same, we can express this equation in terms of the characters of the classes: 1X al ¼ gðcÞwðlÞ ðcÞ wðcÞ ð5:24Þ h c Although the last two expressions provide a formal procedure for finding the reduction coefficients, in many cases it is possible to find them by inspection. For example, in the s-orbital basis for C3v, the characters are (4, 1, 2) for the classes (E, 2C3, 3sv). By inspection of the character table (Table 5.5), it is immediately clear that the reduction is 2A1 þ E. However, in more complicated cases, the formal procedure is almost essential.
5.12 SYMMETRY-ADAPTED BASES
j
147
Example 5.8 How to determine the reduction of a representation
What symmetry species do the four H1s-orbitals of methane span? Method. Methane belongs to the point group Td; the character table can be
found in Appendix 1. The character of each operation in the four-dimensional basis (Ha, Hb, Hc, Hd) can be determined by noting the number (N) of members left in their original location after the application of each operation: a 1 occurs in the diagonal of the representative in each case, and so the character is the sum of 1 taken N times. (If the member of the basis moves, a zero appears along the diagonal which makes no contribution to the character.) Only one operation from each class need be considered because the characters are the same for all members of a class. With the characters w(c) established, apply eqn 5.24 to determine the reduction. C2, S4
b
d
a
Answer. Refer to Fig. 5.25. The numbers of unchanged basis members under the operations E, C3, C2, S4, sd are 4, 1, 0, 0, 2, respectively. The order of the group is h ¼ 24. It follows from eqn 5.24 that
aðA1 Þ ¼ C3
c
aðA2 Þ ¼ aðEÞ ¼ aðT1 Þ ¼ aðT2 Þ ¼
1 24 fð4 1Þ þ 8ð1 1Þ þ 3ð0 1Þ þ 6ð0 1Þ þ 6ð2 1Þg 1 24 fð4 1Þ þ 8ð1 1Þ þ 3ð0 1Þ 6ð0 1Þ 6ð2 1Þg 1 24 fð4 2Þ 8ð1 1Þ þ 3ð0 2Þ þ 6ð0 0Þ þ 6ð2 0Þg 1 24 fð4 3Þ þ 8ð1 0Þ 3ð0 1Þ þ 6ð0 1Þ 6ð2 1Þg 1 24 fð4 3Þ þ 8ð1 0Þ 3ð0 1Þ 6ð0 1Þ þ 6ð2 1Þg
¼1 ¼0 ¼0 ¼0 ¼1
Hence, the four orbitals span A1 þ T2. d Fig. 5.25 The symmetry elements of
the group Td used in Example 5.8.
Comment. In some cases, an operation changes the sign of a member of the
basis without moving its location (an example is the O2px-orbital in H2O under the operation C2). This sign reversal results in 1 appearing on the diagonal. In other cases, such as for the basis (px, py) on the central atom in a molecule belonging to the group C3v, a fractional value appears on the diagonal: see Section 5.13. Self-test 5.8. What symmetry species do the five Cl3s-orbitals of PCl5, a
trigonal bipyramidal molecule in the gas phase, span?
5.12 Symmetry-adapted bases We now establish how to find the linear combinations of the members of a basis that span an irreducible representation of a given symmetry species. This procedure is called finding a symmetry-adapted basis and the resulting basis functions are called symmetry-adapted linear combinations. The next couple of pages will bristle with subscripts; if you do not wish to pick your way through the thicket, you will be able to use the final result (eqn 5.32). We need to define a projection operator: d X ðlÞ ðlÞ D ðRÞ R ð5:25Þ Pij ¼ l h R ij This operator can be thought of as a mixture of the operations of the group, with a weight given by the value of the matrix elements of the representation.
148
j
5 GROUP THEORY
We prove below that the effect of the projection operator is as follows: ðlÞ ðl0 Þ
Pij fj0
ðlÞ
¼ fi dll0 djj0
ð5:26Þ
Proof 5.6 The effect of a projection operator 0
0
0
0
Consider the set of functions f (l ) ¼ (f1(l ), f2(l ), . . . , fd(l ) ) that form a basis for a 0 0 dl 0 -dimensional irreducible representation D(l ) of symmetry species G(l ) of a group of order h. We can express the effect of any operation of the group as X ðl0 Þ ðl0 Þ ðl0 Þ fi0 Di0 j0 ðRÞ Rfj0 ¼ i0
The GOT may now be invoked. First we multiply by the complex conjugate of an element D(l) ij (R) of a representative of the same operation, and then sum over the elements, using the GOT to simplify the outcome: X ðlÞ X X ðlÞ ðl0 Þ ðl0 Þ ðl0 Þ Dij ðRÞ Rfj0 ¼ Dij ðRÞ fi0 Di0 j0 ðRÞ R
R
¼
X
i0 ðl0 Þ fi0
i0
Basis set C
B
D
E
(l )
Pii
Symmetryadapted basis i
(l' )
0
Pji'
(l )
Pii
Symmetryadapted basis i
Pji(l )
j
ðlÞ ðl0 Þ Dij ðRÞ Di0 j0 ðRÞ
)
R
which is equivalent to eqns 5.25 and 5.26.
The reason why P is called a projection operator can now be made clear. In the first case, suppose that either l 6¼ l 0 or j 6¼ j 0 ; then when P(l) ij 0 acts on some member fj(l0 ), it gives zero. That is, when P(l) ij acts on a function that is not a member of the basis set that spans G(l), or—if it is a member—is not at the location j in the set, then it gives zero. On the other hand, if the member is at the location j of the set that does span G(l), then it converts the function standing at the location j into the function standing at the location i. That is, P projects a member from one location to another location (Fig. 5.26). The importance of this result is that if we know only one member of a basis of a representation, then we can project all the other members out of it. In the special case of l 0 ¼ l and i ¼ j, the effect of the projection operator on some member of the basis is ðlÞ ðlÞ
ðlÞ
Pii fj0 ¼ fi dij0
Fig. 5.26 A schematic diagram to
illustrate the effect of the various projection operators.
0
X
ðl Þ h d 0 dii0 djj0 fi0 ¼ dl0 ll i0
h h ðl0 Þ ðlÞ dll0 djj0 fi ¼ d 0 djj0 fi ¼ dl0 dl ll X
A
(
ð5:27Þ
That is, P then either generates 0 (if i 6¼ j 0 ) or regenerates the original function (if i ¼ j 0 ). The significance of this special case will be apparent soon. Now suppose that we are given a linearly independent but otherwise arbitrary set of functions f ¼ (f1, f2, . . . ). An example might be the s-orbital
5.12 SYMMETRY-ADAPTED BASES
If a basis set g 0 ¼ (g10 , g20 , . . . ) is a linear combination of another basis set g ¼ (g1, g2, . . . ) in the form g 0 ¼ gc (as in Proof 5.2), then g can be expressed as a linear combination of g 0 via g ¼ g 0 c1.
j
149
basis we considered earlier. What is the effect of the projection operator P(l) ii on any one member? Just as any member of the symmetry-adapted basis f 0 can be expressed as the appropriate linear combination of the members of the arbitrary basis f, we can express any fj as a linear combination 0 of all the fj(l0 ): X ðl0 Þ fj ¼ f j0 ð5:28Þ l0 ;j0 0
(The expansion coefficients have been absorbed into the fj(l0 ).) If we now operate on eqn 5.28 with the projection operator Pii(l), we obtain X ðlÞ ðl0 Þ X ðlÞ ðl0 Þ ðlÞ Pii fj ¼ Pii fj0 ¼ dll0 dij0 fj0 ¼ fi ð5:29Þ l0 ;j0
l0 ;j0
That is, when P(l) ii operates on any member of the arbitrary initial basis, it generates the ith member of the basis for the irreducible representation of symmetry species G(l). With that member obtained, we can act on it with P(l) ji to construct the jth member of the set. This solves the problem of finding a symmetry-adapted basis. The problem with the method detailed above is that to set up the projection operators we need to know the elements of all the representatives of the irreducible representation. It is normally the case that only the characters (the sums of the diagonal elements) are available. However, even that limited information can be useful. Consider the projection operator p(l) formed by summing P(l) over its diagonal elements: pðlÞ ¼
Basis set
i
C
B
A
X
D
ðlÞ
Pii ¼
dl X ðlÞ D ðRÞ R h i;R ii
ð5:30Þ
E
The sum over the diagonal elements of a representative is the character of the corresponding operation, so pðlÞ ¼ p Symmetryadapted basis
+
+
+
+
dl X ðlÞ w ðRÞ R h R
This operator can therefore be constructed from the character tables alone. Its effect is to generate a sum of the members of a basis spanning an irreducible representation (Fig. 5.27): X ðlÞ X ðlÞ pðlÞ fj ¼ Pii fj ¼ fi ð5:32Þ i
Fig. 5.27 The projection operator p generates a sum of the symmetryadapted basis functions when it is applied to any member of the original basis.
ð5:31Þ
i
The fact that a sum is generated is of no consequence for one-dimensional irreducible representations because in such cases there is only one member of the basis set. However, for two- and higher-dimensional irreducible representations the projection operator gives a sum of two or more members of the basis. Nevertheless, because we are generally concerned only with low-dimensional irreducible representations, this is rarely a severe complication, and the following example shows how any ambiguity can be resolved.
150
j
5 GROUP THEORY
Example 5.9 How to use projection operators
Construct the symmetry-adapated bases for the group C3v using the s-orbital basis. Method. We have already established that the s-orbital basis spans 2A1 þ E, so
we can use eqn 5.32 to construct the appropriate symmetry-adapted bases by projection. We shall take all the characters to be real. The simplest way to use eqn 5.32 is to follow this recipe: 1. Draw up a table headed by the basis and show in the columns the effect of the operations. (A given column is headed by fj and an entry in the table shows Rfj.) 2. Multiply each member of the column by the character of the corresponding operation. (This step produces w(R)Rfj at each location; the characters in Table 5.5 are real.) P 3. Add the entries within each column. (This produces R w(R)Rfj for a given fj.) 4. Multiply by dimension/order. (This produces pfj.) For the group C3v, h ¼ 6. Answer. The table to construct is as follows:
Original set:
sN
sA
sB
sC
Under E
sN
sA
sB
sC
C3þ
sN
sB
sC
sA
C 3
sN
sC
sA
sB
v
sN
sA
sC
sB
v0
sN
sB
sA
sC
v00
sN
sC
sB
sA
For the irreducible representation of symmetry species A1, d ¼ 1 and all w(R) ¼ 1. Hence, the first column gives 1 6 ðsN
þ sN þ sN þ sN þ sN þ sN Þ ¼ sN
The second column gives 1 6 ðsA
þ sB þ sC þ sA þ sB þ sC Þ ¼ 13 ðsA þ sB þ sC Þ
The remaining two columns give the same outcome. For E, d ¼ 2 and for the six operations w ¼ (2,1,1, 0, 0, 0) for the six operations. The first column gives 2 6 ð2sN
sN sN þ 0 þ 0 þ 0Þ ¼ 0
The second column gives 2 6 ð2sA
sB sC þ 0 þ 0 þ 0Þ ¼ 13 ð2sA sB sC Þ
The remaining columns produce 13 ð2sB sC sA Þ and 13 ð2sC sA sB Þ: These three linear combinations are not linearly independent (the sum of them
5.13 THE TRANSFORMATION OF p-ORBITALS
j
151
is zero), so we can form a linear combination of the second two combinations that is orthogonal to the first. The combination s3 ¼ 13 ð2sB sC sA Þ 13 ð2sC sA sB Þ ¼ sB sC is orthogonal to s2 ¼ 13 ð2sA sB sC Þ. Note that the two linear combinations s2 and s3 have a different character under sv (þ1 and 1, respectively). Self-test 5.9. Find the symmetry-adapted linear combinations of the p-orbitals
in NO2.
y
The symmetry properties of functions
y x
x
y y –x
x
y –1/2x + 1/2√3y
We now turn to a consideration of the transformation properties of functions in general. To set the scene, we shall investigate how the three p-orbitals of the nitrogen atom in NH3 transform under the operations of the group C3v. The basis set for the representation we shall develop is (px, py, pz). Intuitively, we can expect the representation to reduce to an irreducible representation spanned by pz because pz ! pz under all operations of the group (but is it of symmetry species A1 or A2?) and a two-dimensional irreducible representation spanned by (px,py) of symmetry species E, because these orbitals are mixed by the symmetry operations. But suppose the basis was extended to include d-orbitals on the central atom—what irreducible representations would then be spanned? To answer questions like that, we need a systematic procedure that can be applied even when—especially when—the conclusions are not obvious. The systematic approach is set out below. The procedures are essentially the same as we have already described, but they are more generally applicable than the calculations done above.
5.13 The transformation of p-orbitals x
–1/2y – 1/2√3x
Consider the basis (px,py,pz) for C3v. We know from Section 3.13 that the orbitals have the form px ¼ xf ðrÞ py ¼ yf ðrÞ pz ¼ zf ðrÞ
y
x –1/2x – 1/2√3y
Fig. 5.28 The effect of certain
symmetry operations of the group C3v on the functions x and y.
where r is the distance from the nucleus. All operations of a point group leave r unchanged, and so the orbitals transform in the same way as the basis (x, y, z). Some of the transformations of this basis are illustrated in Fig. 5.28. The effect of sv on the basis is 2 3 1 0 0 sv ðx, y, zÞ ¼ ðx, y, zÞ ¼ ðx, y, zÞ4 0 1 0 5 0 0 1 we have This relation identifies D(sv) in this basis. Under the rotation Cþ 2 1 3 1pffiffiffi 3 2 2 3 0 pffiffiffi pffiffiffi p ffiffiffi þ C3 ðx, y, zÞ ¼ ð12x þ 12 3y, 12 3x 12y, zÞ ¼ ðx, y, zÞ4 1 3 05 12 2 0 0 1
152
j
5 GROUP THEORY
Table 5.7 The matrix representation of C3v in the basis
(x,y,z)
2
DðEÞ
3
1 0 0 40 1 05 0 0 1 wðEÞ ¼ 3
2
DðCþ DðC 3Þ 3Þ pffiffiffi pffiffiffi 3 2 3 1 1 1 2 2 3 0 2 3 0 p ffiffiffi 12 0 5 4 12 3 12 0 5 0 0 1 0 0 1 wðCþ wðC 3Þ ¼ 0 3Þ¼ 0
12 4 1 pffiffiffi 2 3
Dðs0v Þ Dðs00v Þ p ffiffiffi pffiffiffi 3 2 3 1 1 1 1 1 0 0 2pffiffiffi 2 3 0 2 ffiffiffi 2 3 0 p 4 0 1 05 41 3 12 0 5 4 12 3 12 0 5 2 0 0 1 0 0 1 0 0 1 wðsv Þ ¼ 1 wðs0v Þ ¼ 1 wðs00v Þ ¼ 1 2
Dðsv Þ
3 2
and we can identify D(Cþ 3 ) for the basis. The complete representation can be established in this way, and is set out in Table 5.7, together with the characters. The characters of the operations E, 2C3, and sv in the basis (x, y, z) are 3, 0, and 1, respectively. This corresponds to the reduction A1 þ E. The function z is a basis for A1, and the pair (x,y) span E. We therefore now also know that the three p-orbitals also span A1 þ E, and that pz is a basis for A1 and (px, py) is a basis for E. The identities of the symmetry species of the irreducible representations spanned by x, y, and z are so important that they are normally given explicitly in the character tables (see Appendix 1). Exactly the same procedure may be applied to the quadratic forms x2, xy, etc. that arise when the d-orbitals are expressed in Cartesian coordinates (Section 3.13): dxy ¼ xyf ðrÞ dyz ¼ yzf ðrÞ dzx ¼ zxf ðrÞ dx2 y2 ¼ ðx2 y2 Þf ðrÞ dz2 ¼ ð3z2 r2 Þf ðrÞ and the symmetry species these functions span are also normally reported: in C3v the five functions span A1 þ 2E.
5.14 The decomposition of direct-product bases The question that now arises is stimulated by noticing that the quadratic forms that govern the symmetry properties of the d-orbitals are expressed as products of the linear terms that govern the symmetry properties of p-orbitals. We can now explore whether it is possible to find the symmetry species of quadratic forms such as xy, for instance, directly from the properties of x and y without having to go through the business of setting up the symmetry transformations and their representatives all over again. In more general terms, if we know what symmetry species are spanned by a basis (f1, f2, . . . ), can we state the symmetry species spanned by their products, such as (f12, f1f2, . . . )? We shall now show that this information is carried by the character tables.
5.14 THE DECOMPOSITION OF DIRECT-PRODUCT BASES
j
153
First, we show that if fi(l) is a member of a basis for an irreducible repres0 entation of symmetry species G(l) of dimension dl, and fi(l0 ) is a member of a 0 basis for an irreducible representation of symmetry species G(l ) of dimension dl 0 , then the products also form a basis for a representation, which is called a direct-product representation. Its dimension is dldl 0 . Proof 5.7 The direct-product representation
Under an operation R of a group the two basis functions transform as follows: X ðlÞ ðlÞ X ðl0 Þ ðl0 Þ ðlÞ ðl0 Þ fj Dji ðRÞ Rfi0 ¼ fj0 Dj0 i0 ðRÞ Rfi ¼ j0
j
It follows that their product transforms as X ðlÞ ðl0 Þ ðlÞ ðl0 Þ ðlÞ ðl0 Þ Rfi0 ¼ fj fj0 Dji ðRÞDj0 i0 ðRÞ Rfi j;j0 0
which is a linear combination of the products fj(l)fj(l0 ).
To discover whether the direct-product representation is reducible, we need to work out its characters. The matrix representative of the operation R 0 in the direct-product basis is Dji(l)(R)Dj(l0 i 0)(R), where the pair of indices jj 0 now label the row of the matrix and the indices ii 0 label the column. The diagonal elements are the elements with j ¼ i and j 0 ¼ i 0 . It follows that the character of the operation R is ( )( ) X ðlÞ X ðlÞ X ðl0 Þ ðl0 Þ wðrÞ ¼ Dii ðRÞDi0 i0 ðRÞ ¼ Dii ðRÞ Di0 i0 ðRÞ i;i0 ðlÞ
i
i0
ðl0 Þ
¼ w ðRÞw ðRÞ
ð5:33Þ
This is a very simple and useful result: it states that the characters of the operations in the direct-product basis are the products of the corresponding characters for the original bases. With the characters of the representation established, we can then use the standard techniques described above to decide on the reduction of the representation. This procedure is illustrated in the following example. Example 5.10 The reduction of a direct-product representation
Determine the symmetry species of the irreducible representations spanned by (a) the quadratic forms x2, y2, z2 and (b) the basis (xz, yz) in the group C3v. Method. For both parts of the problem we use the result set out in eqn 5.33 to
establish the characters of the direct-product representation, and then reconstruct that set of characters as a linear combination of the characters of the irreducible representations of the group. If the decomposition of the characters is not obvious, use the procedure set out in Example 5.8. Answer. (a) The basis (x,y,z) spans a (reducible) representation with characters 3, 0, 1 (in the usual order E, 2C3, 3sv). The direct-product basis
154
j
5 GROUP THEORY
composed of x2, y2, z2 therefore spans a representation with characters 9, 0, 1. This set of characters corresponds to 2A1 þ A2 þ 3E. (b) The basis (xz, yz) is the direct product of the bases z and (x, y) which span A1 and E, respectively. The direct-product basis therefore has characters (in the usual order) ð1 1 1Þ ð2 1 0Þ ¼ ð2 1 0Þ which we recognize as the characters of E itself. Therefore, (xz, yz) is a basis for E, as indicated in Appendix 1. Comment. The fact that the direct product of bases that span A1 and E spans E
is normally written A1 E ¼ E Self-test 5.10. What irreducible representations are spanned by the direct
product of (x, y) with itself in the group C3v? [A1 þ A2 þ E]
In the example we have shown that A1 E ¼ E, which is a formal way of expressing the fact that the direct-product basis (xz, yz) spans E. In the same way, the direct product of (x, y) with itself, which consists of the basis (x2, xy, yx, y2), spans E E ¼ A1 þ A2 þ E (The significance of the appearance of both xy and yx is discussed below.) Tables of decompositions of direct products like these are called directproduct tables. They can be worked out once and for all, and some are listed in Appendix 1. We shall see that they are often as important as the character tables themselves! A particularly important point to note from the tables 0 is that the product G(l) G(l ) contains the totally symmetric irreducible representation (A1 in many groups) only if l 0 ¼ l. Finally, we need to account for the presence of both xy and yx in the directproduct basis. We need to note that the symmetrized direct product ðþÞ
fij
ðlÞ ðlÞ
ðlÞ ðlÞ
¼ 12 ffi fj þ fj fi g
ð5:34Þ
and the antisymmetrized direct product ðÞ
fij
ðlÞ ðlÞ
ðlÞ ðlÞ
¼ 12 ffi fj fj fi g
ð5:35Þ
of a basis taken with itself also form bases for the group. Clearly, the latter (eqn 5.35) vanishes identically in this case because xy yx ¼ 0. We need to establish which irreducible representations are spanned by the antisymmetrized direct product and discard them from the decomposition. The characters of the products (eqns 5.34 and 5.35) are given by the following expressions:5 wþ ðRÞ ¼ 12 fwðlÞ ðRÞ2 þ wðlÞ ðR2 Þg
w ðRÞ ¼ 12 fwðlÞ ðRÞ2 wðlÞ ðR2 Þg
ð5:36Þ
.......................................................................................................
5. For a derivation, see M. Hamermesh, Group theory and its applications to physical problems, Addison-Wesley, Reading, Mass. (1962).
j
5.15 DIRECT-PRODUCT GROUPS
155
In the direct-product tables the symmetry species of the antisymmetrized product is denoted [G]. The fact that it is reported at all signifies that it has some use: we shall see what it is in Section 7.16. In the present case E E ¼ A1 þ ½A2 þ E and so we now know that (x2, xy, y2) spans A1 þ E. One of the most important applications of this type of procedure is in the determination of selection rules (see below, Section 5.16).
5.15 Direct-product groups We can now consider another example of using group theory to build up information from existing results. Here we shall show how to build up the properties of larger groups by cementing together the character tables for smaller groups. Suppose there exists a group G of order h with elements R1, R2, . . . , Rh and another group G 0 of order h with elements R10 , R20 , . . . , R0h0 . Let the groups satisfy the following two conditions: 1. The only element in common is the identity. 2. The elements of group G commute with the elements of group G 0 . Because commutation holds, RR 0 ¼ R 0 R. Examples of two such groups are Cs and C3v. Then the products RR 0 of each element of G with each element of G 0 form a group called the direct-product group: G00 ¼ G G0
ð5:37Þ
00
That G is in fact a group can be verified by checking that the group property is obeyed for all pairs of elements. Then, because RiRj ¼ Rk (because G is a group) and Rr0 Rs0 ¼ Rt0 (for a similar reason), in G00 with elements RiRr0 : ðRi R0r ÞðRj R0s Þ ¼ Ri R0r Rj R0s ¼ Ri Rj R0r R0s ¼ Rk R0t and the element so generated is a member of G00 . The order of the directproduct group is hh 0 (so the order of Cs C3v is 2 6 ¼ 12). The direct-product group can be identified by constructing its elements (Cs C3v will turn out to be D3h), and the character table can be constructed from the character tables of the component groups. To do so, we proceed as follows. Let (f1, f2, . . . ) be a basis for an irreducible representation of G and (f10 , f20 , . . . ) be a basis for an irreducible representation of G 0 . It follows that we can write X X fj Dji ðRÞ R0 fr0 ¼ fs0 Dsr ðR0 Þ ð5:38Þ Rfi ¼ s
j
Then the effect of RR 0 on the direct-product basis is X RR0 fi fr0 ¼ ðRfi ÞðR0 fr0 Þ ¼ fj fs0 Dji ðRÞDsr ðR0 Þ j;s
The character of the operation RR 0 is the sum of the diagonal elements: X wðRR0 Þ ¼ Dii ðRÞDrr ðR0 Þ ¼wðRÞwðR0 Þ ð5:39Þ ir
156
j
5 GROUP THEORY
Therefore, the character table of the direct-product group can be written down simply by multiplying together the appropriate characters of the two contributing groups. Example 5.11 How to construct the character table of a direct-product group
Construct the direct-product group Cs C3v, identify it, and build its character table from the constituent groups. Method. To construct the direct-product group, we form elements by com-
bining each element of one group with each element of the other group in turn. It is often sufficient to deal with the products of classes of operation rather than each individual operation. The resulting group is recognized by noting its composition and referring to Fig. 5.16. The characters are constructed by multiplying together the characters contributing to each operation.
h
C 3+ Fig. 5.29 A combination of the
operations sh and Cþ 3 is equivalent to the operation Sþ 3.
h
Answer. The groups Cs and C3v have, respectively, two and three classes, so the direct-product group has 2 3 ¼ 6 classes. It follows that it also has six symmetry species of irreducible representations. The classes of Cs are (E,sh) and those of C3v are (E, 2C3, 3sv). When each class of C3v is multiplied by the identity operation of Cs, the same three classes, (E, 2C3, 3sv), are reproduced. Each of these classes is also multiplied by sh. The operation Esh is the same as þ sh itself. The operations Cþ 3 sh and C3 sh are the improper rotations S3 and S3 , respectively (see Fig. 5.29). The operations svsh are the same as two-fold rotations about the bisectors of the angles of the triangular object (Fig. 5.30) and are denoted C2. The direct-product group is therefore formed as follows: E
C 3v:
C2
v
C s:
Fig. 5.30 A combination of the
operations sh and sv is equivalent to the operation C2.
Cs
E
h
0
1
1
A00
1
1
A
E
2C 3
h
C 3v ⊗ C s: E
E
h
3σv
h
2C 3
E
2S3
3v
h
3C 2
According to the system of nomenclature described in Section 5.2, this set of operations corresponds to the group D3h. At this point, we use the rule about characters to construct the character table. The two component group character tables are shown here and in the margin on p. 145. Upon taking all the appropriate products we obtain the following table:
A10 ( ¼ A1A 0 ) A100 ( ¼ A2A00 ) A20 ( ¼ A2A 0 ) A200 ( ¼ A1A00 ) E 0 ( ¼ EA 0 ) E00 ( ¼ EA00 )
E ¼ EE
sh ¼ Esh
2C3 ¼ E(2C3)
2S3 ¼ sh(2C3)
3sv ¼ E(3sv)
3C2 ¼ sh(3sv)
1 1 1 1 2 2
1 1 1 1 2 2
1 1 1 1 1 1
1 1 1 1 1 1
1 1 1 1 0 0
1 1 1 1 0 0
This is the table for this group given in Appendix 1.
5.16 VANISHING INTEGRALS
f
j
157
Comment. The procedure described here is an important and easy way of constructing the character tables for more complex groups, such as D6h ¼ D6 Ci and Oh ¼ O Ci.
–a
a
0
x
Self-test 5.11. Construct the character table for the group D6h ¼ D6 Ci.
(a)
5.16 Vanishing integrals f
–a
0
a
x
0
a
x
(b)
f
–a (c)
Fig. 5.31 (a) An antisymmetric
function with necessarily zero integral over a symmetric range about the origin. (b) A symmetric function with non-zero integral over a symmetric range. (c) The integral of this symmetric function, however, is zero.
–a a
Fig. 5.32 The symmetry element of a symmetric integration range.
One of the more important applications of group theory is to the problem of deciding when integrals are necessarily zero on account of the symmetry of the system. This application can be illustrated quite simply by considering two functions f(x) and g(x), and the integral over a symmetrical range around x ¼ 0. Let f(x) be a function that is antisymmetric with respect to the interchange of x and x, so f(x) ¼ f(x). The integral of this function over a range from x ¼ a to x ¼ þa is zero (Fig. 5.31). On the other hand, if g(x) is a symmetrical function in the sense that g(x) ¼ g(x), then its integral over the same range is not necessarily zero. Note that the integral of g may, by accident, be zero, whereas the integral of f is necessarily zero. Now consider another way of looking at the two functions. The range (a, a) is considered an ‘object’ with two symmetry elements: the identity and a mirror plane perpendicular to the x-axis (Fig. 5.32). Such an object belongs to the point group Cs. The function f spans the irreducible representation of symmetry species A00 because Ef ¼ f and shf ¼ f. On the other hand, g spans A 0 because Eg ¼ g and sh g ¼ g. That is, if the integrand is not a basis for the totally symmetric irreducible representation of the group, then the integral is necessarily zero. If the integrand is a basis for the totally symmetric irreducible representation, then the integral is not necessarily zero (but may accidentally be zero). This simple example also introduces a further point that generalizes to all groups. The integrals of f 2 and g2 are not zero, but the integral of fg is necessarily zero. This feature is consistent with the discussion above, because f 2 is a basis for A00 A00 ¼ A 0 , which is the totally symmetric irreducible representation; likewise g2 is a basis for A 0 A 0 ¼ A 0 , which is also the totally symmetric irreducible representation. However fg is a basis for A00 A 0 ¼ A00 , which is not totally symmetric, so the integral necessarily vanishes. Another way of looking at this result is to note that f spans one species of irreducible representation, g spans another. Then, basis functions that span irreducible representations of different symmetry species are orthogonal. More formally: if fi(l) is the ith member of a basis that spans the irreducible representation of symmetry species G(l) of a group, and fj(l) is the jth member of a basis that spans the irreducible representation of symmetry species G(l) of the same group, then for a symmetric range of integration: Z ðlÞ ðl0 Þ fi fj dt / dll0 dij ð5:40Þ
158
j
5 GROUP THEORY
The proof of this result is based on the GOT, and is given in Further information 14. Note that the integral may be zero even when l 0 ¼ l and i ¼ j, because the eqn 5.40 is silent concerning the value of the proportionality constant. We have now arrived at one of the most important results of group theory. The conclusion can be summarized as follows: R 0 An integral f (l) f (l ) dt over a symmetric range is necessarily zero unless the integrand is a basis for the totally symmetric irreducible representation of 0 the group which will be the case only if G(l) ¼ G(l ). Example 5.12 The identification of zero integrals
Determine which orbitals of nitrogen in ammonia may have non-vanishing overlap with the symmetry-adapted linear combinations s1, s2, and s3 of hydrogen 1s-orbitals specified in Example 5.4. R Method. The overlap integral has the form ci cj dt; hence it is non-vanishing only if Gi Gj includes A1. Begin by identifying the symmetry species of the N2s- and N2p-orbitals by using the character table in Appendix 1 and noting that px transforms as x, etc., and decide which can have non-vanishing overlap with the symmetry-adapted linear combinations of the H1s-orbitals. Use the direct-product tables in Appendix 1. Recall from Section 5.9 that s1 spans the irreducible representation A1 and (s2, s3) spans E. Answer. In C3v, the N2p-orbitals span A1(pz) and E(px,py). Because A1 A1 ¼ A1 and E E ¼ A1 þ A2 þ E, the N2pz orbital can have non-zero overlap with the combination s1, and the px and py orbitals can have non-zero overlap with s2 and s3. The N2s-orbital also spans A1, and so may also overlap with s1. Comment. Note that whether the s1 symmetry-adapted linear combination
has non-zero overlap with N2pz depends on the bond angle: when the molecule is flat, s1 lies in the nodal plane of N2pz and the overlap is zero. Self-test 5.12. Show using group theory that the overlap of s1 and N2pz is
necessarily zero when the molecule is planar.
An integral of the form Z 0 00 I ¼ f ðlÞ f ðl Þ f ðl Þ dt
ð5:41Þ
over all space is also necessarily zero unless the integrand is a basis for the totally symmetric irreducible representation (such as A1). To determine 0 whether that is so, we first form G(l) G(l ) and expand it in the normal way. (k) Then we take each G in the expansion and form the direct product 00 G(k) G(l ). If A1 (or the equivalent totally symmetric irreducible representation) occurs nowhere in the resulting expression, then the integral I is necessarily zero. In other words, the integral I necessarily vanishes if the 00 symmetry species G(l ) does not match one of the symmetry species in the
5.17 SYMMETRY AND DEGENERACY
j
159
0
direct product G(l) G(l ). This conclusion is of the greatest importance in quantum mechanics because we often encounter integrals of the form Z hajOjbi ¼ ca Ocb dt Therefore, we can use group theory to decide when matrix elements are necessarily zero. This often results in an immense simplification of the construction of molecular orbitals, the interpretation of spectra, and the calculation of molecular properties. Example 5.13 The identification of vanishing matrix elements
Do the integrals (a) hdxy jzj dx2 y2 i and (b) hdxy jlz j dx2 y2 i vanish in a C4v molecule? 0
00
Method. We need to assess whether G(l) G(l ) G(l ) contains A1. To do so,
we use the character tables in Appendix 1 to identify the symmetry species of each function in the integral. Angular momenta transform as rotations (Section 5.18) so lz transforms as the rotation Rz, which is listed in the tables. Use Appendix 1 for the direct-product decomposition. Answer. In C4v, dxy and dx2 y2 span B2 and B1, respectively, whereas z spans A1 and lz spans A2. (a) The integrand spans
B2 A1 B1 ¼ B2 B1 ¼ A2 and hence the matrix element must vanish. (b) The integrand spans B2 A2 B1 ¼ B2 B2 ¼ A1 and hence the integral is not necessarily zero. Comment. Matrix elements of this kind are particularly important for
discussing electronic spectra: we shall see that they occur in the formulation of selection rules. Self-test 5.13. Does the integral hdxy j lz j dxzi vanish in a C3v molecule?
5.17 Symmetry and degeneracy We have already mentioned (in Section 2.15) that the presence of degeneracy is a consequence of the symmetry of a system. We are now in a position to discuss this relation. To do so, we note that the hamiltonian of a system must be invariant under every operation of the relevant point group: ðRHÞ ¼ H
ð5:42Þ
A qualitative interpretation of eqn 5.42 is that the hamiltonian is the operator for the energy, and energy does not change under a symmetry operation. An example is the hamiltonian for the harmonic oscillator: the kinetic energy operator is proportional to d2/dx2 and the potential energy operator is proportional to x2. Both terms are invariant under the replacement of x by x, and so the hamiltonian spans the totally symmetric irreducible representation
160
j
5 GROUP THEORY
of the point group Cs. Because H is invariant under a similarity transformation of the group (that is, any symmetry operation leaves it unchanged), we can write RHR1 ¼ H Multiplication from the right by R gives RH ¼ HR, so we can conclude that symmetry operations must commute with the hamiltonian. We now demonstrate that functions that can be generated from one another by any symmetry operation of the system have the same energy. That is: Eigenfunctions that are related by symmetry transformations of the system are degenerate. We have already seen an example of this result in the discussion of the geometrically square two-dimensional square-well eigenfunctions in Section 2.15. Proof 5.8 Degeneracy and symmetry
Consider an eigenfunction ci of H with eigenvalue E. That is, Hci ¼ Eci. We can multiply this equation from the left by R, giving RHci ¼ ERci, and insert R1R for the identity, to obtain RHR1 Rci ¼ ERci From the invariance of H it then follows that HRci ¼ ERci Therefore, ci and Rci correspond to the same energy E.
We can go on to formulate a rule for the maximum degree of degeneracy that can occur in a system of given symmetry. Consider a member cj of a basis for an irreducible representation of dimension d of the point group for the system, and suppose it has an energy E. We have already seen that all the other members of the basis can be generated by acting on this function with the projection operator Pij defined in eqn 5.25. However, because Pij is a linear combination of the symmetry operations of the group, it commutes with the hamiltonian. Therefore, Pij Hcj ¼ HPij cj ¼ Hci
and
Pij Hcj ¼ Pij Ecj ¼ Eci
and hence Hci ¼ Eci, and ci has the same eigenvalue as cj. But we can generate all d members of the d-dimensional basis by choosing the index i appropriately, and so all d basis functions have the same energy. We can conclude that: The degree of degeneracy of a set of functions is equal to the dimension of the irreducible representation they span. This dimension is always given by w(E), the character of the identity. In the harmonic oscillator, with point group Cs, the only irreducible representations are one-dimensional, and therefore all the eigenfunctions are non-degenerate. For a geometrically square two-dimensional square-well
5.18 THE GENERATORS OF ROTATIONS
j
161
potential, with point group C4v, two-dimensional irreducible representations are allowed, and so some levels can be doubly degenerate. Triply degenerate levels occur in systems with cubic point-group symmetry, and five-fold degeneracy is encountered in icosahedral systems. The full rotation group, R3, has irreducible representations of arbitrarily high dimension, so degeneracies of any degree can occur.
The full rotation group We shall now consider the full rotation groups in two and three dimensions (R2 and R3) and discover the deep connection between group theory and the quantum mechanics of angular momentum. The techniques are no different in principle from those introduced earlier in the chapter, but there are some interesting points of detail.
5.18 The generators of rotations y y
x
x
y
y – x
x +y x
Consider first the full rotation group R2 in two dimensions, the point group of a circular system (Fig. 5.33). The name R2 is a synonym of C1v and is an example of an infinite rotation group in the sense that rotations through any angles (and in particular infinitesimal angles) are symmetry operations. You should bear in mind the analogous illustration for the equilateral triangle (Fig. 5.28) to see the similarities and differences between finite and infinite rotation groups. We shall first establish the effect of an infinitesimal counter-clockwise rotation through an angle df about the z-axis on the basis (x, y). It will be convenient to work in polar coordinates and to write the basis as (r cos f, r sin f), with r a constant under all operations of the group. Under the infinitesimal rotation df, which we denote Cdf, the basis transforms as follows: Cdf ðx; yÞ ¼ fr cosðf dfÞ, r sinðf dfÞg ¼ fr cos f cos df þ r sin f sin df, r sin f cos df r cos f sin dfg ¼ fr cosf þ r df sin f þ , r sin f r df cos f þ g ¼ ðx þ y df þ , y x df þ Þ ¼ ðx, yÞ ðy, xÞdf þ We have used the expansions sin x ¼ x 16 x3 þ and cos x ¼ 1 12x2 þ and have kept only lowest-order terms in the infinitesimal angle df. That is:
Fig. 5.33 The effect on the functions
x and y of an infinitesimal rotation df about the z-axis.
Cdf ðx, yÞ ¼ ðx, yÞ ðy, xÞdf þ Now we identify an important fact. Consider the effect of the angular momentum operator
h q q ð5:43Þ lz ¼ x y i qy qx
162
j
5 GROUP THEORY
on the basis: lz ðx, yÞ ¼
h q q h x y ðx, yÞ ¼ ðy, xÞ i qy qx i
By comparing this result with the effect of Cdf, we see that i Cdf ðx, yÞ ¼ 1 dflz þ ðx, yÞ h
(x )
C (y )
C
(y )
C
(x )
C
z (x )
(y ) C (x )
C
x
C
(y ) C
y
Fig. 5.34 The non-commutation of
perpendicular rotations. Notice that the outcome of the combined ðyÞ ðxÞ rotation Cdb Cda is different ðxÞ ðyÞ from the outcome of Cda Cdb :
ð5:44Þ
and that the operator itself can be written i ð5:45Þ Cdf ¼ 1 dflz þ h The infinitesimal rotation operator therefore differs from the identity to first order in df by a term that is proportional to the operator lz. The operator 1 (i/ h)dflz is therefore called the generator of the infinitesimal rotation about the z-axis. In a similar way, the operators lx and ly are the generators for rotations about the x- and y-axes in R3. We know that the angular momentum operators satisfy a set of commutation relations. These can be seen in a different light as follows. The effect of a sequence of rotations about different axes depends on the order in which they are applied (Fig. 5.34). Under a rotation by da about x followed by a rotation by db about y, we have
i i ðyÞ ðxÞ Cdb Cda ¼ 1 dbly þ 1 dalx þ h h
2 i i dbdaly lx þ ¼ 1 ðdbly þ dalx Þ þ h h However, if the rotations are applied in the opposite order the outcome is
i i ðxÞ ðyÞ Cda Cdb ¼ 1 dalx þ 1 dbly þ h h
2 i i ¼ 1 ðdbly þ dalx Þ þ dbdalx ly þ h h The difference between these two operations to second order is
2 i i ðyÞ ðxÞ ðxÞ ðyÞ Cdb Cda Cda Cdb ¼ dadbðly lx lx ly Þ ¼ dadblz h h
ð5:46Þ
where the last equality follows from the commutation relation [lx,ly] ¼ ihlz. The result we have established is that the difference between two infinitesimal rotations is equivalent to a single infinitesimal rotation through the angle dadb about the z-axis, which is geometrically plausible (Fig. 5.34). The reverse argument, that it is geometrically obvious that the difference is a single rotation, therefore implies that [lx,ly] ¼ ihlz. Hence, the angular momentum commutation relations can be regarded as a direct consequence of the geometrical properties of composite rotations.
5.19 The representation of the full rotation group We shall now look for the irreducible representations of the full rotation group R3. As a starting point, we note that the spherical harmonics Ylml for
5.19 THE REPRESENTATION OF THE FULL ROTATION GROUP
j
163
a given l transform into linear combinations of one another under a rotation. (For example, p-orbitals rotate into one another, d-orbitals do likewise, and so on, but p-orbitals do not rotate into d-orbitals. This is consistent with the result that eigenfunctions related by symmetry transformations are degenerate.) Therefore, the functions Yll, Yl,l 1, . . . ,Yl, l form a basis for a (2l þ 1)dimensional (and it turns out, irreducible) representation of the group. Each spherical harmonic has the form Ylml ¼ P(y)eimlf, and so, as a result of a rotation by a around the z-axis, each one transforms into P(y)eiml(f a). The entire basis therefore transforms as follows: ilðfaÞ , PðyÞeiðl1ÞðfaÞ , . . . , PðyÞeilðfaÞ CðzÞ a ðYll , Yl;l1 , . . . , Yl;l Þ ¼ PðyÞe 3 2 ila 0 0 0 e 6 0 eiðl1Þa 0 0 7 7 6 6 .. 7 7 6 ð5:47Þ ¼ ðYll , Yl;l1 , . . . , Yl;l Þ6 0 0 . 7 7 6 .. 7 .. 6 .. 4 . . 5 . 0
0
eila
This expression lets us recognize the matrix representative of the rotation in the basis. The character of a rotation through the angle a about the z-axis (and therefore about any axis, because in R3 all rotations through a given angle belong to the same class) is the following sum: wðCa Þ ¼ eila þ eiðl1Þa þ þ eila ¼ 1 þ 2 cos a þ 2 cos 2a þ þ 2 cosðl 1Þa þ 2 cos la ix
The sum of a series a þ ar þ ar 2 þ þ ar n is a(r n þ 1 1)/(r 1). In the present case, a ¼ eila, r ¼ eia, and n ¼ 2l. To evaluate (sin ax)/(sin bx) in the limit x ! 0, recall that in this same limit (sin ax)/(ax) ¼ 1, and therefore (sin ax)/(sin bx) ¼ a/b. In the present case, x ¼ a, a ¼ l þ 12, and b ¼ 12.
ð5:48Þ
ix
To obtain this expression, we have used e þ e ¼ 2 cos x; the leading 1 comes from the term with ml ¼ 0. This simple expression can be used to establish the character of any rotation for a (2l þ 1)-dimensional basis. An even simpler version is obtained by recognizing that the first line is a geometric series. Hence, it is the sum wðCa Þ ¼
l X
eiml a ¼
ml ¼l
eila ðeið2lþ1Þa 1Þ eia 1
ð5:49Þ
This slightly awkward expression can be manipulated into wðCa Þ ¼
sinðl þ 12Þa sin 12 a
ð5:50Þ
In the limit a ! 0 pertaining to an infinitesimal rotation, the character is 2l þ 1, and so the levels with quantum number l are (2l þ 1)-fold degenerate in a spherical system. Example 5.14 How to determine the symmetry species of atoms in various
environments An atom has a configuration that gives rise to a state with l ¼ 3. What symmetry species would it give rise to in an octahedral environment?
164
j
5 GROUP THEORY
Method. We need to identify the rotations that are common to both R3 and O,
and then to calculate their characters from eqn 5.50 with l ¼ 3. Then, by referring to the character table for O in Appendix 1, we can identify the symmetry species spanned by the state in the reduced symmetry environment. Answer. The rotation angles in O (recall in R3 all angles are permitted) are a ¼ 0 for E, a ¼ 2p/3(C3), p(C2), p/2(C4), p(C20 ). Because
wðCa Þ ¼
sinð7a=2Þ sinða=2Þ
we find w ¼ (7, 1, 1, 1, 1) for (E, C3, C2, C4, C20 ). Then, use of eqn 5.24 with h ¼ 24 gives a(A2) ¼ 1, a(T1) ¼ 1, and a(T2) ¼ 1. Therefore, in the reduced symmetry environment the symmetry species are A2 þ T1 þ T2. Comment. The step down from a group to its subgroup is called ‘descent in
symmetry’. It is a particularly important technique in the theory of the structure and spectra of d-metal complexes (see Chapter 8). The atomic configuration with l ¼ 3 is called an F term; the descent in symmetry in this case is denoted F ! A2 þ T1 þ T2. Self-test 5.14. What irreducible representations does an l ¼ 4 state (a G term)
span in tetrahedral symmetry?
5.20 Coupled angular momenta We now explore the group-theoretical description of the coupling of two angular momenta. We suppose that we have two sets of functions that are the bases for irreducible G(j1) and G(j2) of the full rotation group. The functions ðj Þ ðj Þ ðj Þ ðj Þ will be denoted fmj11 and fmj22 , respectively. The products fmj11 fmj22 provide a ðj1 Þ ðj2 Þ basis for the direct-product representation G G . This representation is in general reducible, and we can reduce it as explained in Section 5.14. First, we write X aj GðjÞ ð5:51Þ Gðj1 Þ Gðj2 Þ ¼ j
To determine the coefficients we consider the characters: wðCa Þ ¼ wðj1 Þ ðCa Þwðj2 Þ ðCa Þ ¼
j1 X
j2 X
eiðmj1 þmj2 Þa
ð5:52Þ
mj1 ¼j1 mj2 ¼j2
The question we now address is whether the right-hand side of this equation can be expressed as a sum over Smj eimja and, if so, how many times each term in the sum appears. We shall now demonstrate that each term appears exactly once, and that j varies from j1 þ j2 down to j j1 j2 j . The argument runs as follows. Because j mj1 þ mj2 j j1 þ j2, it follows that j mj j j1 þ j2, and so j j1 þ j2. Therefore, aj ¼ 0 if j > j1 þ j2. The maximum value of mj may be obtained from mj1 and mj2 in only one way: when mj1 ¼ j1
5.20 COUPLED ANGULAR MOMENTA
j
165
and mj2 ¼ j2. Therefore, aj1 þ j2 ¼ 1. The next value of mj, which is j 1, may be obtained in two ways, namely mj1 ¼ j1 1 and mj2 ¼ j2 or mj1 ¼ j1 and mj2 ¼ j2 1; one of these ways is accounted for by the representation with j ¼ j1 þ j2, and so we can conclude that aj1þj21 ¼ 1. This argument can be continued down to j ¼ j j1 j2 j , and so eqn 5.52 is equivalent to wðCa Þ ¼
jX 1 þj2
j X
j¼jj1 j2 j mj ¼j
eimj a ¼
jX 1 þj2
wðjÞ ðCa Þ
ð5:53Þ
j¼jj1 j2 j
Therefore, we can conclude that the direct product decomposes as follows: Gðj1 Þ Gðj2 Þ ¼ Gðj1 þj2 Þ þ Gðj1 þj2 1Þ þ þ Gðjj1 j2 jÞ
ð5:54Þ
which is nothing other than the Clebsch–Gordan series, eqn 4.44. This result shows, in effect, that the whole of angular momentum theory can be regarded as an aspect of group theory and the symmetry properties of rotations.
Applications There are numerous applications of group theory, both explicit and implicit. We shall encounter many of them in the following pages. That being so, we shall only indicate here the types of applications that are encountered, and where in the text. The application of the rotation groups (R3, D1h, and C1v) will appear wherever we discuss the angular momentum of atoms and molecules (Chapters 7, 10, and 11). Finite groups play an important role in the discussion of molecular structure and properties, both in the setting up of molecular orbitals (Chapter 8) and in the evaluation of the matrix elements and expectation values that are needed to evaluate molecular properties (Chapter 6). When an atom or ion is embedded in a local environment, as in a crystal or a complex, the degeneracy of its orbitals is removed with important consequences for its spectroscopic features (Chapter 11). Spectroscopy in general also relies heavily on group-theoretical arguments in its classification of states, the construction of normal modes of vibration, and the derivation of selection rules. The calculation of the electric and magnetic properties of molecules relies on the evaluation of matrix elements, and group theory helps by eliminating many integrals on the basis of symmetry alone (Chapters 12 and 13). The following chapters will confirm that group theory does indeed pervade the whole of quantum chemistry.
166
j
5 GROUP THEORY
PROBLEMS 5.1 Classify the following molecules according to their point symmetry group: (a) H2O, (b) CO2, (c) C2H4, (d) cis-ClHC ¼ CHCl, (e) trans-ClCH ¼ CHCl, (f) benzene, (g) naphthalene, (h) CHClFBr, (i) B(OH)3. 5.2 Which of the molecules listed above may possess a permanent electric dipole moment? Hint. Decide on R the criterion for the non-vanishing of hi ¼ c c dt and refer to the tables in Appendix 1; transforms as r ¼ (x, y, z). 5.3 Find the representatives of the operations of the group C2v using as a basis the valence orbitals of H and O in H2O (that is, H1sA, H1sB, O2s, O2p). Hint. The group is of order 4 and so there are four 6-dimensional matrices to find. 5.4 Confirm that the representatives established in Problem 5.3 reproduce the group multiplications C22 ¼ E, svC2 ¼ sv0 . 5.5 Determine which symmetry species are spanned by the six orbitals of H2O described in Problem 5.3. Find the symmetry-adapted linear combinations, and confirm that the representatives are in block-diagonal form. Hint. Decompose the representation established in Problem 5.3 by analysing the characters. Use the projection operator in eqn 5.31 to establish the symmetry-adapted bases (using the elements of the representatives established in Problem 5.3), form the matrix of coefficients cji (Section 5.6) and use eqn 5.7 to construct the irreducible representations. 5.6 Find the representatives of the operations of the group Td by using as a basis four 1s-orbitals, one at each apex of a regular tetrahedron (as in CH4). Hint. The basis is fourdimensional; the order of the group is 24, and so there are 24 matrices to find. 5.7 Confirm that the representations established in Problem 5.6 reproduce the group multiplications C þ 3C 3 ¼ E, 0 S4C3 ¼ S4, and S4C3 ¼ sd. 5.8 Determine which irreducible representations are spanned by the four 1s-orbitals in methane. Find the symmetry-adapted linear combinations, and confirm that the representatives for Cþ3 and S4 are in block-diagonal form. Hint. Decompose the representation into irreducible representations by analysing the characters. Use the projection operator in eqn 5.31 to establish the symmetryadapted bases. 5.9 Analyse the following direct products into the symmetry species they span: (a) C2v: A2 B1 B2, (b) C3v: A1 A2 E, (c) C6v: B2 E1, (d) C1v: E12, (e) O: T1 T2 E.
5.10 Show that 3x2y y3 is a basis for an A1 irreducible 2 3 representation of C3v. Hint. Show that Cþ 3 (3x y y ) / 3x2y y3; likewise for the other elements of the group. 5.11 A function f(x,y,z) was found to be a basis for a representation of C2v, the characters being (4, 0, 0, 0). What symmetry species of irreducible representations does it span? Hint. Proceed by inspection to find the al in eqn 5.21 or use eqn 5.23. 5.12 Find the components of the function f(x,y,z) (from Problem 5.11) acting as a basis for each irreducible representation it spans. Hint. Use eqn 5.31. The basis for A1, for example, turns out to be 14{f(x, y, z) þ f(x, y, z) þ f(x, y, z) þ f(x, y, z)}. 5.13 Regard the naphthalene molecule as having C2v symmetry (with the C2 axis perpendicular to the plane), which is a subgroup of its full symmetry group. Consider the p-orbitals on each carbon as a basis. What symmetry species do they span? Construct the symmetry-adapted bases. Hint. Proceed as in Example 5.9. 5.14 Repeat the process of Problem 5.13 for benzene, using the subgroup C6v of the full symmetry group. After constructing the symmetry-adapted linear combinations, refer to the D6h character table to label them according to the full group. 5.15 Show that in an octahedral array, hydrogen 1s-orbitals span A1g þ Eg þ T1u of the group Oh. 5.16 Classify the terms that may arise from the following configurations: (a) C2v: a21 b11 b12 ; (b) C3v: a12 e1 , e2 ; (c) Td: a12 e1 , e1 t11 , t11 t21 , t12 , t22 ; (d) O: e2 , e1 t11 , t22 . Hint. Use the direct product tables; triplet terms have antisymmetric spatial functions. 5.17 Construct the character tables for the groups Oh and D6h. Hint. Use D6h ¼ D6 Ci and Oh ¼ O Ci and the procedure in Section 5.15. 5.18 Demonstrate that there are no non-zero integrals of R the form c 0 Hc dt when c 0 and c belong to different symmetry species. 5.19 The ground states of the C2v molecules NO2 and ClO2 are 2A1 and 2B1, respectively; the ground state of O2 is 3S g . To what states may (a) electric-dipole, (b) magnetic-dipole transitions take place? Hint. The electric-dipole operator transforms as translations, the magnetic as rotations. 5.20 What is the maximum degeneracy of the energy levels of a particle confined to the interior of a regular tetrahedron?
PROBLEMS
5.21 Demonstrate that the linear momentum operator p ¼ (h/i)(d/dx) is the generator of infinitesimal translations. Hint. Proceed as in eqn 5.45. 5.22 An atom bearing a single p-electron is trapped in an environment with C3v symmetry. What symmetry species does it span? Hint. Use eqn 5.49 with a ¼ 120 . 5.23 The group multiplication table for C2v is shown in Example 5.2. Confirm that the group elements multiply associatively. 5.24 A molecule of carbon dioxide, initially in a S u electronic state, absorbs z-polarized electromagnetic
j
167
radiation. What is the symmetry of the excited electronic state? 5.25 In the square-planar xenon tetrafluoride molecule, consider the symmetry-adapted linear combination p1 ¼ pA pB þ pC pD where pA, pB, pC, pD are the 2pz atomic orbitals on the fluorine atoms (clockwise labelling of the fluorine atoms). Which of the various s, p, and d atomic orbitals on the central xenon atom can overlap with p1 to form molecular orbitals? Hint: It will be much easier to work in the reduced point group D4 rather than the full symmetry point group of the molecule.
6 Time-independent perturbation theory 6.1 Perturbation of a two-level system 6.2 Many-level systems 6.3 The first-order correction to the energy 6.4 The first-order correction to the wavefunction 6.5 The second-order correction to the energy 6.6 Comments on the perturbation expressions 6.7 The closure approximation 6.8 Perturbation theory for degenerate states Variation theory 6.9 The Rayleigh ratio 6.10 The Rayleigh–Ritz method The Hellmann–Feynman theorem Time-dependent perturbation theory 6.11 The time-dependent behaviour of a two-level system 6.12 The Rabi formula 6.13 Many-level systems: the variation of constants 6.14 The effect of a slowly switched constant perturbation 6.15 The effect of an oscillating perturbation 6.16 Transition rates to continuum states 6.17 The Einstein transition probabilities 6.18 Lifetime and energy uncertainty
Techniques of approximation
This is a sad but necessary chapter. It is sad because we have reached the point at which the hope of finding exact solutions is set aside and we begin to look for methods of approximation. It is necessary, because most of the problems of quantum chemistry cannot be solved exactly, so we must learn how to tackle them. There are very few problems for which the Schro¨dinger equation can be solved exactly, and the examples in previous chapters almost exhaust the list. As soon as the shape of the potential is distorted from the forms already considered, or more than two particles interact with one another (as in a helium atom), the equation cannot be solved exactly. There are three ways of making progress. The first is to try to guess the shape of the wavefunction of the system. Even people with profound insight need a criterion of success, and this is provided by the ‘variation principle’, which we specify below. It is useful to be guided to the form of the wavefunction by a knowledge of the distortion of the system induced by the complicating aspects of the potential or the interactions. For example, the exact solutions for a system that resembles the true system may be known and can be used as a guide to the true solutions by noting how the hamiltonians of the two systems differ. This procedure is the province of ‘perturbation theory’. Perturbation theory is particularly useful when we are interested in the response of atoms and molecules to electric and magnetic fields. When these fields change with time (as in a light wave) we have to deal with ‘timedependent perturbation theory’. The third important method of approximation, which is dealt with in detail in Chapters 7 and 9, makes use of ‘self-consistent field’ procedures, which is an iterative method for solving the Schro¨dinger equation for systems of many particles.
Time-independent perturbation theory In time-independent perturbation theory we make use of the fact that the hamiltonians for the true and simpler model system, H and H(0), respectively, differ by a contribution that is independent of the time: H ¼ Hð0Þ þ Hð1Þ
ð6:1Þ
We refer to H(1) as the perturbation. Our aim is to generate the wavefunctions and energy of the perturbed system from a knowledge of the unperturbed
6.1 PERTURBATION OF A TWO-LEVEL SYSTEM
j
169
system and a procedure for taking into account the presence of the perturbation.
6.1 Perturbation of a two-level system Consider first a system that has only two eigenstates. We suppose that the two eigenstates of H(0) are known, and denote them j1i and j2i. The corresð0Þ ð0Þ ponding wavefunctions are c1 and c2 , respectively. These states and functions form a complete orthonormal basis. They correspond to the enerð0Þ ð0Þ gies E1 and E2 : ð0Þ ð0Þ Hð0Þ cð0Þ m ¼ Em cm
m ¼ 1, 2
The wavefunctions of the true system differ only slightly from those of the model system, and we can hope to solve the equation Hc ¼ Ec
ð6:2Þ
in terms of them by writing ð0Þ
ð0Þ
c ¼ a1 c1 þ a2 c2
ð6:3Þ
where a1 and a2 are constants to be determined. To find the constants am we insert the linear combination into the Schro¨dinger equation and obtain (using ket notation) a1 ðH EÞj1i þ a2 ðH EÞj2i ¼ 0 When this equation is multiplied from the left by the bras h1j and h2j in turn, and use is made of the orthonormality of the two states, we obtain the two equations a1 ðH11 EÞ þ a2 H12 ¼ 0 a1 H21 þ a2 ðH22 EÞ ¼ 0
ð6:4Þ
where Hmn ¼ hmjHjni. The condition for the existence of non-trivial solutions of this pair of equations is that the determinant of the coefficients of the constants a1 and a2 should disappear (see Further Information 23 and Example 1.10): H11 E H12 ¼0 H21 H22 E This expression expands to ðH11 EÞðH22 EÞ H12 H21 ¼ 0 and then to E2 ðH11 þ H22 ÞE þ H11 H22 H12 H21 ¼ 0 This quadratic equation has the solutions n o1=2 E ¼ 12ðH11 þ H22 Þ 12 ðH11 H22 Þ2 þ 4H12 H21
ð6:5Þ
170
j
6 TECHNIQUES OF APPROXIMATION
In the special case of a perturbation for which the diagonal matrix elements ð1Þ ð0Þ are zero (Hmm ¼ 0, so we can write Hmm ¼ Em ), this expression simplifies to 1=2 ð0Þ ð0Þ ð0Þ ð0Þ 2 2 1 1 E ¼ 2 E1 þ E2 2 E1 E2 þ 4e ð6:6Þ
2
E 2 + /∆ E E– Energy
E2
ð1Þ
∆E E1 E+ E1 – 2/∆E
Fig. 6.1 The variation of the energies
of a two-level system with a constant perturbation as the separation of the unperturbed levels is increased. The pale lines show the energies according to second-order perturbation theory.
ð1Þ
ð1Þ
where e2 ¼ H12 H21 . Because H(1) is hermitian, we can write e2 ¼ jH12 j2 . ð0Þ ð0Þ When the perturbation is absent, e ¼ 0 and Eþ ¼ E1 , E ¼ E2 , the two unperturbed energies. Figure 6.1 shows the variation of the energies of the system as the separation of the states of the model system is increased. As can be seen, the lower of the two levels is lowered in energy and that of the upper level is raised. In other words, the effect of the perturbation is to drive the energy levels apart and to prevent their crossing. This non-crossing rule is a common feature of all perturbations that can link two states (that is, for which ð1Þ Hmn 6¼ 0 for m 6¼ n). A second general feature can also be seen from the illustration: the effect of the perturbation is greater the smaller the energy separation of the unperturbed levels. For instance, when the two original ð0Þ ð0Þ energies have the same energy (E1 ¼ E2 ), then Eþ E ¼ 2e Equation 6.6 also shows that the stronger the perturbation, the stronger the effective repulsion of the levels. In summary:
2
/∆E
Energy
(a)
(b)
2
∆E
2/∆E
Fig. 6.2 (a) When the unperturbed levels are far apart in energy, the shift in energy caused by a perturbation of strength e is e2/DE. (b) If the levels are initially degenerate, then the shift in energy is much larger, and is equal to e.
1. When a perturbation is applied, the lower level moves down in energy and the upper level moves up. 2. The closer the unperturbed states are in energy, the greater the effect of a perturbation. 3. The stronger the perturbation, the greater the effect on the energies of the levels. The effect of the perturbation can be seen in more detail by considering the case of a perturbation that is weak compared with the separation of the ð0Þ ð0Þ energy levels in the sense that e2 ðE1 E2 Þ2 . When this condition holds, we can expand eqn 6.6 by making use of (1 þ x)1/2 ¼ 1 þ 12x þ , to obtain 8 91=2 > < = > 2 4e ð0Þ ð0Þ ð0Þ ð0Þ 1þ E ¼ 12 E1 þ E2 12 E1 E2 2 > > ð0Þ ð0Þ : ; E1 E2 9 8 > = < > 2 2e ð0Þ ð0Þ ð0Þ ð0Þ 1 1 1þ ¼ 2 E1 þ E2 2 E1 E2 2 þ > > ð0Þ ð0Þ ; : E1 E2 from which it follows that to second-order in e we have ð0Þ
Eþ E1
e2 DEð0Þ ð0Þ
ð0Þ
E E2 þ ð0Þ
e2 DEð0Þ
ð6:7Þ
where DEð0Þ ¼ E2 E1 (Fig. 6.2). These two solutions converge on the exact solutions when (2e/DE(0))2 1, as shown in Fig. 6.1. A general feature
6.2 MANY-LEVEL SYSTEMS
j
171
of all perturbation theory calculations is that the shifts in energy are of the order of e2/DE(0). The perturbed wavefunctions are obtained by solving eqn 6.4 for the coefficients setting in turn E ¼ Eþ (to obtain cþ) and E ¼ E (to obtain c). A convenient way to express the solutions is to write ð0Þ
cþ ¼ c01 cos z þ c2 sin z
ð0Þ
ð0Þ
c ¼ c1 sin z þ c2 cos z
ð6:8Þ
for this ensures that cþ and c are orthonormal for all values of z. Then it is found (see Example 1.10 for details) that1 ð1Þ 2H12 tan 2z ¼ ð0Þ ð6:9Þ ð0Þ E1 E2 ð0Þ
ð0Þ
For a degenerate model system ðE1 ¼ E2 Þ, we have tan 2z ¼ 1, corresponding to z ¼ p/4. In this case the perturbed wavefunctions are 1 ð0Þ 1 ð0Þ ð0Þ ð0Þ cþ ¼ 1=2 c1 þ c2 c ¼ 1=2 c2 c1 ð6:10Þ 2 2 It follows that each perturbed state is a 50 per cent mixture of the two model states. In contrast, for a perturbation acting on two widely separated states ð1Þ we can write tan 2z 2z ¼ 2jH12 j=DEð0Þ . Furthermore, because sin z z and cos z 1, it follows from eqn 6.8 that ð1Þ ð1Þ H12 H12 ð0Þ ð0Þ ð0Þ ð0Þ c c c2 þ c ð6:11Þ cþ c1 DEð0Þ 2 DEð0Þ 1 We see that each model state is slightly contaminated by the other state.
6.2 Many-level systems Now we generalize these results to a system in which there are numerous, and possibly an infinite number of, non-degenerate levels. Special precautions have to be taken if the state of interest is degenerate, and we consider that possibility in Section 6.8. We suppose that we know all the eigenfunctions and eigenvalues of a model system with hamiltonian H(0) that differs from the true system to a small extent. An example might be an anharmonic oscillator or a molecule in a weak electric field: the model systems would then be a harmonic oscillator or a molecule in the absence of a field, respectively. We therefore suppose that we have found the solutions of the equations Hð0Þ jni ¼ Eð0Þ n jni
ð6:12Þ
with n ¼ 0, 1, 2, . . . , and jni a member of an orthonormal basis. We shall suppose that we are calculating the perturbed form of the state j0i of energy ð0Þ E0 , but this state is not necessarily the ground state of the system. .......................................................................................................
ð1Þ
ð1Þ
1. In general, a complex matrix element H12 can be written as jH12 jeif . In the following, we suppose that f ¼ 0.
172
j
6 TECHNIQUES OF APPROXIMATION
The hamiltonian of the perturbed system will be written H ¼ Hð0Þ þ lHð1Þ þ l2 Hð2Þ þ
ð6:13Þ
The only significance of the parameter l is that it keeps track of the order of the perturbation, and will enable us to identify all first-order terms in the energy, all second-order terms, and so on. At the end of the calculation we set l ¼ 1 because by then it will have served its purpose. Similarly, the perturbed wavefunction of the system will be written ð0Þ
ð1Þ
ð2Þ
c0 ¼ c0 þ lc0 þ l2 c0 þ
ð6:14Þ ð0Þ
which shows how the unperturbed function ðc0 Þ is corrected by terms that are of various orders in the perturbation. The energy of the perturbed state also has correction terms of various orders, and we write ð0Þ
ð1Þ
ð2Þ
E0 ¼ E0 þ lE0 þ l2 E0 þ
ð6:15Þ
ð1Þ
ð2Þ
We shall refer to E0 as the first-order correction to the energy, to E0 as the second-order correction, and so on. The equation to solve is Hc ¼ Ec
ð6:16Þ
Insertion of the preceding equations into this equation, followed by collecting terms that have the same power of l, then results in ð0Þ
ð0Þ
ð0Þ
l0 fHð0Þ c0 E0 c0 g ð1Þ
ð0Þ
ð2Þ
ð1Þ
ð0Þ
ð1Þ
ð1Þ
ð0Þ
ð0Þ
ð0Þ
ð2Þ
þ l1 fHð0Þ c0 þ Hð1Þ c0 E0 c0 E0 c0 g ð1Þ
ð1Þ
þ l2 fHð0Þ c0 þ Hð1Þ c0 þ Hð2Þ c0 E0 c0 E0 c0 ð2Þ
ð0Þ
E0 c0 g þ ¼ 0 Because l is an arbitrary parameter, the coefficient of each power of l must equal zero separately, so we have the following set of equations: ð0Þ
ð0Þ
ð0Þ
Hð0Þ c0 ¼ E0 c0 ð0Þ
ð1Þ
ð1Þ
ð0Þ
ð0Þ ð2Þ E0 gc0
ð2Þ fE0
ð0Þ Hð2Þ gc0
fHð0Þ E0 gc0 ¼ fE0 Hð1Þ gc0 fHð0Þ
¼
ð6:17Þ þ
ð1Þ fE0
ð1Þ Hð1Þ gc0
and so on.
6.3 The first-order correction to the energy The solution of the first of eqn 6.17 is assumed known (it is eqn 6.12). The first-order correction to the wavefunction is written as a linear combination of the unperturbed wavefunctions of the system because the latter constitute a complete basis set of functions: X ð1Þ c0 ¼ an cð0Þ ð6:18Þ n n
6.3 THE FIRST-ORDER CORRECTION TO THE ENERGY
j
173
The sum is over all states of the model system including those belonging to the continuum, if there is one. When this expansion is inserted into the equation ð1Þ for c0 , we obtain (in ket notation) o o X n X n ð0Þ ð0Þ jni an Hð0Þ E0 jni ¼ an Eð0Þ n E0 n
n
n o ð1Þ ¼ E0 Hð1Þ j0i
ð6:19Þ
When this expression is multiplied from the left by the bra h0j we obtain o n o X n ð0Þ ð1Þ an Eð0Þ h0 j ni ¼ h0j E0 Hð1Þ j0i n E0 n ð1Þ
¼ E0 h0jHð1Þ j0i The left-hand side of this equation is zero, so we can conclude that the firstorder correction to the energy of the state j0i is ð1Þ
ð1Þ
E0 ¼ h0jHð1Þ j0i ¼ H00
ð6:20Þ
ð1Þ H00
The matrix element is the average value of the first-order perturbation over the unperturbed state j0i. An analogy is the first-order shift in the frequency of a violin string when small weights are added along its length: those at the nodes have no effect on the frequency, those at the antinodes (the points of maximum amplitude) affect the frequency most strongly, and the overall effect is an average taking into account the displacement of the string at the location of each weight. In the special case in which the diagonal matrix elements of the perturbation are zero, there is no first-order correction to the energy. How to calculate the first-order correction to the energy
Potential energy, V
Example 6.1
A small step in the potential energy is introduced into the one-dimensional square-well problem (Fig. 6.3). Calculate the first-order correction to the energy of a particle confined to the well and evaluate it for a ¼ L/10, so the blip in the potential occupies the central 10 per cent of the well, and for (a) n ¼ 1, (b) n ¼ 2. a
Method. We need to evaluate eqn 6.20 by using
0
L/2
L
Fig. 6.3 The perturbation to a square-well potential used in Example 6.1.
We have used the integral Z sin2 kx dx ¼ 12x
1 sin 2kx þ constant 4k
x
Hð1Þ ¼
e if 12 ðL aÞ x 12 ðL þ aÞ 0 if x is outside this region
The wavefunctions are given in eqn 2.31. We should anticipate that the effect of the perturbation will be much smaller for n ¼ 2 than for n ¼ 1 because in the former the perturbation is applied in the vicinity of a node. Answer. The integral required is
Eð1Þ n ¼
2e L
Z
1 2ðLþaÞ
1 2ðLaÞ
sin2
npx a ð1Þn npa sin dx ¼ e np L L L
With a ¼ L/10, (a) for n ¼ 1, E(1) ¼ 0.1984e; (b) for n ¼ 2, E(1) ¼ 0.0065e.
174
j
6 TECHNIQUES OF APPROXIMATION
Comment. The relative sizes of the two answers are consistent with the per-
turbation being close to an antinode and a node, respectively. When n is very large, E(1) (a/L)e, independent of n. At such high quantum numbers, the probability of finding the particle in the region a is a/L regardless of n. Note that if e > 0, then the energy of the states is increased from the unperturbed values. Self-test 6.1. Evaluate the first-order correction to the energy of a particle in a
box for a perturbation of the form e sin(xp/L) for n ¼ 1 and n ¼ 2.
6.4 The first-order correction to the wavefunction Now we look for the first-order correction to the state of the system. To find it, multiply eqn 6.19 from the left by the bra hkj, where k 6¼ 0: o E D n o E X D n ð1Þ ð0Þ an k Eð0Þ n ¼ k E0 Hð1Þ 0 n E0 n
The orthonormality of the states again simplifies this expression to n o ð0Þ ð0Þ ð1Þ ak Ek E0 ¼ E0 hk j 0i hkjHð1Þ j0i ¼ hkjHð1Þ j0i ð0Þ
ð0Þ
Because the state j0i is non-degenerate, the differences Ek E0 non-zero for k 6¼ 0. Therefore, the coefficients are given by
are all
ð1Þ
ak ¼
Hk0 ð0Þ
ð0Þ
E0 Ek
ð6:21Þ
ð1Þ
where Hk0 ¼ hkjHð1Þ j0i. It follows that the wavefunction of the system corrected to first-order in the perturbation is ( ) ð1Þ X Hk0 ð0Þ ð0Þ 0 ck ð6:22Þ c0 c0 þ ð0Þ ð0Þ E E k 0 k where the prime on the sum means that the state with k ¼ 0 should be omitted. The last equation echoes the expression derived for the two-level system in the limit of a weak perturbation and widely separated energy levels. As in that case, perturbation theory guides us towards the form of the perturbed state of the system. In this case, the procedure simulates the distortion of the state by mixing into it the other states of the system. This mixing is expressed by saying that the perturbation induces virtual transitions to these other states of the model system. However, that is only a pictorial way of speaking: in fact, the distorted state is being simulated as a linear superposition of the unperturbed states of the system. The equation shows that a particular state k ð1Þ makes no contribution to the superposition if Hk0 ¼ 0, and (for a given magnitude of the matrix element) the contribution of a state is smaller the ð0Þ ð0Þ larger the energy difference jE0 Ek j.
6.5 THE SECOND-ORDER CORRECTION TO THE ENERGY
j
175
6.5 The second-order correction to the energy We use the same technique to extract the second-order correction to the energy from eqn 6.17. First, we write the second-order correction to the wavefunction as the linear combination X ð2Þ bn cð0Þ ð6:23Þ c0 ¼ n n
and then substitute this expansion into the third equation in eqn 6.17, which in ket notation becomes o E n o E X n o E X n ð0Þ ð2Þ ð1Þ bn Eð0Þ an E0 Hð1Þ n n ¼ E0 Hð2Þ 0 þ n E0 n
n
Now multiply this equation through from the left by h0j, which gives o D n o E X n ð2Þ ð0Þ ð2Þ E 0 j n ¼ 0 bn Eð0Þ E H h i 0 n 0 0 n
þ
X n
ð2Þ ¼ E0
D n o E ð1Þ an 0 E0 Hð1Þ n
h0jHð2Þ j0i þ
X n
D n o E ð1Þ an 0 E0 Hð1Þ n
n o ð1Þ h0jH j0i þ a0 E0 h0jHð1Þ j0i o E X D n ð1Þ 0 an 0 E0 Hð1Þ n þ
ð2Þ ¼ E0
ð2Þ
n
The left-hand side is zero, as is (from eqn 6.20) the third term on the right as ð1Þ well as the term E0 h0jni in the final sum (because n 6¼ 0), so X ð2Þ 0 an h0jHð1Þ jni E0 ¼ h0jHð2Þ j0i þ n
At this point we can import eqn 6.21 for the coefficients an, and obtain the following expression for the second-order correction to the energy: ð2Þ
ð2Þ
E0 ¼ H00 þ
X Hð1Þ Hð1Þ 0 0n n0 n
ð0Þ
ð0Þ
E0 En
ð6:24Þ
As usual, the prime on the sum signifies omission of the state with n ¼ 0. Equation 6.24 is very important and we shall use it frequently. It is a generalization of the approximate form of the solutions for the two-level ð2Þ problem, and consists of two parts. One, H00 , is the same kind of average as occurs for the first-order correction, and is an average of the secondorder perturbation over the unperturbed wavefunction of the system. The second term is more involved, but can be interpreted as the average of the first-order perturbation taking into account the first-order distortion of the original wavefunction. It should be noticed that because by hermiticity ð1Þ ð1Þ ð1Þ ð1Þ ð1Þ H0n Hn0 ¼ H0n H0n ¼ jH0n j2 , the sum in eqn 6.24 gives a negative conð0Þ ð0Þ tribution (lowers the energy) if En >E0 for all n, which is the case if j0i is the ground state.
176
j
6 TECHNIQUES OF APPROXIMATION
How to evaluate a second-order correction to the energy
Example 6.2
Potential energy, V
Suppose that the square-well potential was modified by the addition of a contribution of the form e sin(px/L) (Fig. 6.4). Find the second-order correction to the energy of the state with n ¼ 1 (the ground state, in this problem) by numerical evaluation of the perturbation sum. ð1Þ
Method. Evaluate the matrix elements Hn0 (where the ‘0’ state of interest here
0 0
L
x
is the ground state with quantum number 1) analytically using the wavefunctions given in eqn 2.31. The denominator in eqn 6.24 is obtained from the energy expression in eqn 2.31 and is proportional to 1 n2. Evaluate the terms in the perturbation sum using mathematical software. By symmetry, only odd ð1Þ values of n contribute. In this problem, H(2) ¼ 0 and Hn0 is real. Answer. The matrix elements we require are as follows:
Fig. 6.4 The perturbation to a
square-well potential used in Example 6.2.
ð1Þ
ð1Þ
Z 2e L npx px px sin sin dx sin L 0 L L L e 1 1 1 ¼ fð1Þn 1g p n 2ðn þ 2Þ 2ðn 2Þ
Hn0 ! Hn1 ¼
h2 ð0Þ ð0Þ ð0Þ 2 E0 Eð0Þ n ! E1 En ¼ 1 n 8mL2 We must therefore evaluate the following sum (where the sum starts at n ¼ 3 because the lowest value of n, n ¼ 1, is omitted and all terms with n even are zero): 2 32mL2 e2 X 1 1 1 1 ð2Þ ð2Þ E0 ! E1 ¼ n 2ðn þ 2Þ 2ðn 2Þ h2 p2 n¼3;5;... 1 n2 ¼
32mL2 e2 8:953 103 2 2 h p
Comment. The distorted wavefunction can be calculated from eqn 6.22 and is ð0Þ
c1 ¼ c1
o 8mL2 e n ð0Þ ð0Þ ð0Þ 2:12c3 þ 0:101c5 þ 0:0168c7 þ 2 100h
This wavefunction corresponds to a greater accumulation of amplitude in the middle of the well. Self-test 6.2. Repeat the calculation for a perturbation of the form
e sin(2px/L).
6.6 Comments on the perturbation expressions We could now go on to find the second-order correction to the wavefunction, and use that result to deduce the third-order correction to the energy, and so on. However, such high-order corrections are only rarely needed and more advanced techniques are generally employed. Furthermore, a useful theorem states that to know the energy correct to order 2n þ 1 in the perturbation, it is sufficient to know the wavefunctions only to nth order in the perturbation. Thus, from the first-order wavefunction, we can calculate the energy up to third order. A final technical problem is to know whether the perturbation
6.6 COMMENTS ON THE PERTURBATION EXPRESSIONS
j
177
theory expansion actually converges. This is answered affirmatively for most common cases by a theorem due to Rellich and Kato,2 but it is normally simply assumed that convergence occurs. The Further reading section suggests places where this delicate question can be pursued. The practical difficulty with eqn 6.24 is that we do not normally have detailed information about the states and energies that occur in the sum. The sum extends, for instance, over all the states of the system, which includes the continuum, if that exists. There are, happily, several aspects of the formulation that diminish this problem. In the first place, the contribution of states that differ by a large energy from the state of interest can be expected to be small on account of the appearance of energy differences in the denominator. Other things being equal, only energetically nearby states contribute appreciably to the sum. The continuum states are generally so high in energy (they correspond, for instance, to ionized states of the system), that they can often safely be ignored. A further apparent difficulty is that although states that are high in energy make only small individual contributions to the sum, there may be very many of them, so their total contribution may be significant. For the hydrogen atom, the number of states of a given energy (that is, the degeneracy) increases as n2, and when n ¼ 103 there are 106 states of the same energy, each one making a small contribution to the sum. However, it often turns out that the matrix elements in the numerators of the perturbation sum vanish identically for many states. For instance, for a hydrogen atom in a uniform electric field in the z-direction, for each n only one of the n2 states of the same energy (the npz-orbital) has non-vanishing matrix elements to the ground state of the atom. Thus, although there may be 106 states lining up to be included, only one of them is selected. The vanishing of matrix elements that so greatly simplifies the perturbation formulas and helps to guarantee convergence of perturbation expansions depends on the symmetry properties of the system. This is where group theory plays such a striking role. The matrix elements of interest are in fact integrals: Z ð1Þ ð0Þ ð6:25Þ H0n ¼ c0 Hð1Þ cð0Þ n dt We saw in Section 5.16 that such integrals are necessarily zero unless the direct product G(0) G(pert) G(n) contains the totally symmetric irreducible representation (for instance, A1 or its equivalent). The physical basis of this important conclusion can be understood by considering the distortion of the wavefunction induced by the perturbation. Suppose that the state of interest (the state j0i) is totally symmetric (it might be the 1s-orbital of a hydrogenic atom). Then Gð0Þ GðpertÞ GðnÞ ¼ A1 GðpertÞ GðnÞ ¼ GðpertÞ GðnÞ and this product must contain A1 (or its equivalent). It does so only if G(pert) ¼ G(n). It follows that the only states that are mixed into the ground state by the perturbation are those with the same symmetry as the perturbation. In other .......................................................................................................
2. See the volume edited by C.H. Wilcox, Perturbation theory and its applications in quantum mechanics, Wiley, New York (1966), for a discussion of these matters.
178
j
6 TECHNIQUES OF APPROXIMATION
words: the distortion impressed on the system has the same symmetry as the perturbation; the perturbation leaves its footprint on the system. For example, if the perturbation is an electric field in the z-direction, then only the pz-orbitals of the atom have the correct symmetry to mirror the effect of the perturbation and are the only orbitals to be included in the sum. Example 6.3
How to determine the states to include in a perturbation
calculation What orbitals should be mixed into a d-orbital when it is perturbed by the application of an electric field in the x-direction? Method. An electric field of strength e in the x-direction corresponds to the
perturbation H(1) ¼ mxe, where mx is the x-component of the electric dipole moment operator: mx ¼ ex. Therefore, we need to decide which matrix elements hdjxjni are non-zero. To do so, we decide on the symmetry species for orbital jni that gives the totally symmetric irreducible representation when we evaluate G(d) G(x) G(n). We use the full rotation group and the results of Section 5.20. In addition, further symmetry analysis can often reduce the list of candidates for the admixed orbitals. Answer. The function for a d-orbital (l ¼ 2) is a component of the basis for G(2)
and x is likewise a component of the basis for G(1) (recall px / x). Because G(2) G(1) ¼ G(3) þ G(2) þ G(1) by eqn 5.54, at this stage we can infer that f-, d-, and p-orbitals can be mixed into the d-orbital. However, under the symmetry operation of inversion, all of the d-orbitals are even but x is odd; therefore the admixed function must be odd, which eliminates d-orbitals. The appropriate functions are therefore f and p. Comment. If a particular d-orbital were specified, only specific f- and
p-orbitals would be in the admixture. For example, of the three p-orbitals, only pz would mix with a dzx-orbital. Self-test 6.3. What orbitals would be mixed into a p-orbital for a field applied
Energy
in the z-direction?
6.7 The closure approximation ∆E
0 Fig. 6.5 The qualitative basis of the closure approximation, in which it is supposed that the individual excitation energies can all be set equal to a single average value.
It is sometimes useful to make a ‘back-of-the-envelope’ assessment of the magnitude of a property without evaluating the perturbation sum in detail. If the spectrum of energy levels of the system resembles that shown in Fig. 6.5, ð0Þ ð0Þ then we can make the approximation that all the energy differences En E0 in the perturbation expression can be replaced by their average value DE. Then the expression for the second-order correction to the energy becomes 1 X 0 ð1Þ ð1Þ ð2Þ ð2Þ H0n Hn0 E0 H00 DE n The sum is almost in the form of a matrix product: X Arn Bnc ¼ ðABÞrc n
6.7 THE CLOSURE APPROXIMATION
j
179
It would be such a product if the sum extended over all n, including n ¼ 0. So, we extend the sum, but cancel the term that should not be present: 1 X ð1Þ ð1Þ 1 ð1Þ ð1Þ ð2Þ ð2Þ H H E0 H00 H0n Hn0 þ DE n DE 00 00 ð6:26Þ 1 ð1Þ ð1Þ 1 ð1Þ ð1Þ ð2Þ H H H00 H00
H00 þ 00 DE DE The energy correction is now expressed solely in terms of integrals over the ground state of the system and we need no information about excited states other than their average energy above the ground state. Because the approximation effectively ‘closes’ the sum over matrix elements down into a single term, it is called the closure approximation. The closure approximation for the second-order energy can be expressed succinctly by introducing the term e2 ¼ h0jHð1Þ2 j0i h0jHð1Þ j0i2
ð6:27aÞ
for then it becomes e2 ð6:27bÞ DE We shall use this expression several times later in the text. Two comments are in order at this point. One is that the closure approximation is a very crude procedure in most instances, because the array of energy levels often differs quite significantly from that supposed in Fig. 6.5. The energy levels of a particle in a box is an example of an array of levels that is quite different from the bunching supposed in the approximation. However, an alternative way of regarding the approximation is to identify DE not with a mean energy but with the following ratio: ð2Þ
ð2Þ
E0 H00
ð1Þ2
H00 ðHð1Þ2 Þ00 . ð0Þ ð0Þ 0 H ð1Þ H ð1Þ E E n 0n n0 0
DE ¼ P n
ð6:28Þ
With this definition of DE, the closure approximation is exact; but of course the net effect is to create more work, and the formal procedure is only useful in so far as it establishes the significance of De somewhat more precisely. Example 6.4
How to use the closure approximation
Derive an approximate expression for the ground-state energy of a hydrogen atom in the presence of an electric field of strength e applied in the z-direction by using the closure approximation. Method. The perturbation hamiltonian is H(1) ¼ mze ¼ eze. The first-order
correction to the energy is zero because eeh0jzj0i ¼ 0. (That the integral vanishes can be easily deduced as follows: the ground state j0i is proportional to Y0,0 and z is proportional to Y1,0 so the symmetry species of the integrand G(0) G(1) G(0) does not include the totally symmetric irrep G(0).) There is no second-order component of the hamiltonian, so the energy expression is
180
j
6 TECHNIQUES OF APPROXIMATION
slightly simplified in so far as it has no terms in H(2). Set up the expression for ð2Þ E0 and then apply closure. The resulting expression can be simplified by taking into account the spherical symmetry of the atom in its ground state and relating the expectation value of z2 to the expectation value of r2. Answer. The full perturbation expression is ð2Þ
E0 ¼ e2 e2
X n
0
z0n zn0 ð0Þ ð0Þ E0 En
We now apply closure, and note that h0jzj0i ¼ 0 by symmetry; therefore, from eqn 6.27a, e2 ¼ e2 e2 h0jz2 j0i The expectation value of z2 in a spherical system is the same as the expectation values of x2 and y2, and because r2 ¼ x2 þ y2 þ z2 it follows that h0jz2 j0i ¼ 13h0jr2 j0i ¼ 13hr2 i where hr2i is the mean square radius of the atom in its ground state. It follows from eqn 6.27b that e2 e2 r2 ð2Þ E0 3DE (a)
Comment. This is a very much simpler expression than the full perturbation
formula. The mean excitation energy may be identified with the ionization energy of the atom, which is close to hcRH, where RH is the Rydberg constant for the hydrogen atom (see Section 7.1). Self-test 6.4. Derive a similar expression for the effect of an electric field on a Perturbed wavefunctions
one-dimensional harmonic oscillator treated as an electric dipole of magnitude ex and force constant k.
6.8 Perturbation theory for degenerate states (b)
Fig. 6.6 A representation of the importance of making the correct choice of basis when considering the effect of a perturbation on degenerate states. In this diagram, the perturbation is represented by the squashing of the circle in a vertical direction. (a) A good choice of basis, because the wavefunctions undergo least change. (b) A poor choice, because both linear combinations are extensively distorted by the perturbation.
Figure 6.1 warns us that the totally wrong result may be obtained for systems in which perturbations are applied to degenerate states, because the ð0Þ ð0Þ denominators En E0 then stand the risk of becoming zero. Another problem with degeneracies is that a small perturbation can induce very large changes in the forms of functions. This point is illustrated schematically in Fig. 6.6, where we see that the perturbation (the effect of which is represented by the conversion of a circle to an ellipse) leads to a large change in the initial pair of degenerate states for one particular choice of starting functions, but to a much more modest change for another choice in which the nodes remain in the same locations. The fact that any linear combination of degenerate functions is also an eigenfunction of the hamiltonian means that we have the freedom to select the combination that most closely resembles the final form of the functions once the perturbation has been applied. We shall now show that both these problems—the selection of optimum starting combinations and the avoidance of zeros in the energy denominators—can be solved by a single procedure.
6.8 PERTURBATION THEORY FOR DEGENERATE STATES
j
181
We suppose that the energy level of interest in the system is r-fold degenð0Þ erate and that the states corresponding to the energy E0 are j0, li, with l ¼ ð0Þ 1, 2, . . . , r; the corresponding wavefunctions are c0;l . All r states satisfy ð0Þ
Hð0Þ j0, li ¼ E0 j0, li
ð6:29Þ
The linear combinations of the degenerate states that most closely resemble the perturbed states are ð0Þ
f0;i ¼
r X
ð0Þ
cil c0;l
ð6:30Þ
i¼1 ð0Þ
When the perturbation is applied, the state f0;i is distorted into ci, which it ð0Þ closely resembles, and its energy changes from E0 to Ei, which has a similar value. The index i is needed on the new energy Ei because the degeneracy may be removed by the perturbation. As in Section 6.2, we write ð0Þ
ð1Þ
ð0Þ
ð1Þ
ci ¼ f0;i þ lc0;i þ Ei ¼ E0 þ lE0;i þ Substitution of these expansions into Hci ¼ Eici and collection of powers of l, just as for the non-degenerate case, gives (up to first order in l) ð0Þ
ð0Þ
ð0Þ
Hð0Þ f0;i ¼ E0 f0;i ð0Þ
ð1Þ
ð1Þ
ð0Þ
fHð0Þ E0 gc0;i ¼ fE0;i Hð1Þ gf0;i
ð6:31Þ
As before, we attempt to express the first-order correction to the wavefunction as a sum over all functions. The simplest procedure is to divide the sum into two parts, one being a sum over the members of the degenerate set j0,li, and the other the sum over all the other states (which may or may not have degeneracies among themselves): X X ð1Þ ð0Þ 0 al c0;l þ an cð0Þ c0;i ¼ n n
l
On insertion of this expression into eqn 6.31 and conversion to ket notation, we obtain o o X n X n ð0Þ ð0Þ ð0Þ 0 jni al E0 E0 j0, li þ an Eð0Þ n E0 l
¼
X
cil
n
ð1Þ E0;i
H
ð1Þ
o
n
j0, li
l
The first term is zero. On multiplying the remaining terms from the left by the bra h0,kj, we obtain zero on the left (because the states jni are orthogonal to the states j0,ki), and hence we are left with o X n ð1Þ cil E0;i h0,kj0, li h0, kjHð1Þ j0, li ¼ 0 l
The degenerate functions need not be orthogonal, so we introduce the following overlap integral: Skl ¼ h0, kj0, li
ð6:32Þ
182
j
6 TECHNIQUES OF APPROXIMATION
If the degenerate functions are orthogonal, the overlap integral Skl ¼ dkl. Similarly, we write ð1Þ
Hkl ¼ h0, kjHð1Þ j0, li
ð6:33Þ
Then we obtain X
n o ð1Þ ð1Þ cil E0;i Skl Hkl ¼ 0
ð6:34Þ
l
These equations (there is one for each value of i) are called the secular equations. They are a set of r simultaneous equations for the coefficients cil and have non-trivial solutions only if the secular determinant is equal to zero: ð1Þ
ð1Þ
detjHkl E0;i Skl j ¼ 0
ð6:35Þ ð1Þ
The solution of this equation gives the energies E0;i that we seek. The solution of the secular equations for each of these values of the energy then gives the coefficients that define the optimum form of the linear combinations to use for any subsequent perturbation distortion. The perturbation of degenerate states
Example 6.5
What is the first-order correction to the energies of a doubly degenerate pair of orthonormal states? Method. We set up the secular determinant and solve it for the energies by
expanding it and looking for the roots of the resulting polynomial in E. Because the pair of states is orthonormal, Skl ¼ dkl. Answer. The secular determinant is
ð1Þ H Eð1Þ 11 0;i ð1Þ H21
ð1Þ ð1Þ ¼ 0 H E ð1Þ
H12
22
0;i
This equation expands to ð1Þ
ð1Þ
ð1Þ
ð1Þ
ð1Þ
ð1Þ
ðH11 E0;i ÞðH22 E0;i Þ H12 H12 ¼ 0 which corresponds to the following quadratic equation for the energy: ð1Þ2
ð1Þ
ð1Þ
ð1Þ
ð1Þ
ð1Þ
ð1Þ
ð1Þ
E0;i ðH11 þ H22 ÞE0;i þ ðH11 H22 H12 H21 Þ ¼ 0 The roots of this equation are ð1Þ
ð1Þ
ð1Þ
ð1Þ
ð1Þ
E0;i ¼ 12fH11 þ H22 g 12fðH11 þ H22 Þ2 ð1Þ
ð1Þ
ð1Þ
ð1Þ
4ðH11 H22 H12 H21 Þg1=2 Comment. This result is the same as we obtained for the two-level problem in
Section 6.1.
6.9 THE RAYLEIGH RATIO
j
183
Variation theory Another very useful method for estimating the energy and approximating the wavefunction of a known hamiltonian is based on variation theory. Variation theory is a way of assessing and improving guesses about the forms of wavefunctions in complicated systems. The first step is to guess the form of a trial function, ctrial, and then the procedure shows how to optimize it.
6.9 The Rayleigh ratio We suppose that the system is described by a hamiltonian H, and denote the lowest eigenvalue of this hamiltonian as E0. The Rayleigh ratio, e, is then defined as R c Hctrial dt e ¼ R trial ð6:36Þ ctrial ctrial dt Then the variation theorem states that for any ctrial , e E0
ð6:37Þ
The equality holds only if the trial function is identical to the true groundstate wavefunction of the system. Proof 6.1 The variation theorem
The trial function can be written as a linear combination of the true (but unknown) eigenfunctions of the hamiltonian (which form a complete set): X ctrial ¼ cn cn where Hcn ¼ En cn n
Now consider the integral Z I ¼ ctrial ðH E0 Þctrial dt Z X cn cn0 cn ðH E0 Þcn0 dt ¼ n;n0
¼
X
cn cn0 ðEn0 E0 Þ
Z
cn cn0 dt
n;n0
¼
X
cn cn ðEn E0 Þ 0
n
The final inequality follows from En E0 and jcnj2 0. It follows that Z ctrial ðH E0 Þctrial dt 0 which rearranges into e E0.
The significance of the variation theorem is that the trial function giving the lowest Rayleigh ratio is the optimum function of that form. Moreover,
j
6 TECHNIQUES OF APPROXIMATION
because the Rayleigh ratio is not less than the true ground-state energy of the system, we have a way of calculating an upper bound to the true energy of the system. Typically, the trial function is expressed in terms of one or more parameters that are varied until the Rayleigh ratio is minimized (Fig. 6.7). The procedure is illustrated in the following example.
Rayleigh ratio
184
Example 6.6
Param
eter p
1
am
er et
p2
r Pa Optimum parameters
Fig. 6.7 The variation principle seeks the values of the parameters (two are shown here) that minimize the energy. The resulting wavefunction is the optimum wavefunction of the selected form.
Using the variation theorem to find an optimized wavefunction
Find the optimum form of a trial function of the form ekr and the upper bound to the ground-state energy of a hydrogenic atom. Method. Begin by writing the hamiltonian for the problem and then evaluate
the integrals that occur in the expression for the Rayleigh ratio. The ratio will be obtained as a function of the parameter k, so to find the minimum value of the ratio we need to find the value of k that corresponds to de/dk ¼ 0. Answer. The hamiltonian for the atom is
H¼
2 2 h Ze2 r 2m 4pe0 r
However, because the trial function is independent of angle, we need consider only the radial derivatives in the laplacian (see eqn 3.18): 1 d2 rc r dr2 The integrals we require are therefore r2 c ¼
3
2p
Z
1=ð4k Þ 2 zfflfflfflffl}|fflfflfflffl{ zfflfflfflfflfflfflffl Z p ffl}|fflfflfflfflfflfflfflffl{ zfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflffl{ Z 1 Z 2p p ctrial ctrial dt ¼ df sin y dy e2kr r2 dr ¼ 3 k 0 0 0 1=ð4k2 Þ
zfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflffl{ Z 2p Z p Z 1 1 p ctrial df sin y dy e2kr rdr ¼ 2 ctrial dt ¼ r k 0 0 0 ! Z Z 1 d2 2 ctrial r ctrial dt ¼ ctrial rekr dt r dr2 Z 2k ctrial dt ¼ ctrial k2 r Z Z 1 c dt ¼ k2 ctrial ctrial dt 2k ctrial r trial p 2p p ¼ ¼ k k k
Z
Therefore, Z p h2 Ze2 ctrial Hctrial dt ¼ 2mk 4e0 k2 and the Rayleigh ratio is e¼
ðph2 =2mkÞ ðZe2 =4e0 k2 Þ k2 h2 Ze2 k ¼ 3 2m p=k 4pe0
j
6.10 THE RAYLEIGH–RITZ METHOD
185
This function is plotted in Fig. 6.8. To find its minimum value we differentiate with respect to k:
5
0
de kh2 Ze2 Ze2 m ¼ ¼ 0 when k ¼ dk m 4pe0 4pe0 h2
a=4
The best value of e is therefore a=5
–5
e¼
0
1
2
k
3
32p2 e20 h2
and the optimum form of the wavefunction has the value of k given above.
a=6 –10
Z2 e4 m
4
Fig. 6.8 The function derived in Example 6.6, with a ¼ Ze2m/2pe0 h2. Note that the minimum is found at k ¼ a/2.
5
Comment. This optimum value of the Rayleigh ratio turns out to be the exact
ground-state energy and the corresponding trial function is the true wavefunction for the atom. This special result follows from the fact that the trial function happens to include the exact wavefunction as a special case. 2
Self-test 6.6. Repeat the calculation for a trial function of the form ekr and
confirm that the Rayleigh ratio lies above the true energy of the ground state.
6.10 The Rayleigh–Ritz method The variation procedure we have described was devised by Lord Rayleigh. A modification called the Rayleigh–Ritz method represents the trial function by a linear combination of fixed basis functions with variable coefficients; these coefficients are treated as the variables to be changed until an optimized set is obtained. The trial function is taken to be X ci c i ð6:38Þ ctrial ¼ i
with only the coefficients (not the basis functions ci) variable; we shall suppose that all coefficients and basis functions are real. The Rayleigh ratio is R P P ci cj ci Hcj dt ci cj Hij R ctrial Hctrial dt i;j i;j R e¼ R ð6:39Þ ¼ P ¼ P ci cj Sij ci cj ci cj dt ctrial ctrial dt i;j
i;j
To find the minimum value of this ratio, we differentiate with respect to each coefficient in turn and set qe/qck ¼ 0 in each case: P P P P P cj Hkj þ ci Hik cj Skj þ ci Sik ci cj Hij qe j i j i i;j P ¼ P 2 qck ci cj Sij c c S i j ij i;j i;j
P P cj Hkj eSkj ci ðHik eSik Þ j P ¼ þ i P ¼0 ci cj Sij ci cj Sij i;j
i;j
186
j
6 TECHNIQUES OF APPROXIMATION
This expression is satisfied if the numerators vanish, which means that we must solve the secular equations X ci ðHik eSik Þ ¼ 0 ð6:40Þ i
This is a set of simultaneous equations for the coefficients ci. The condition for the existence of solutions is that the secular determinant should be zero: detjHik eSik j ¼ 0
ð6:41Þ
Solution of eqn 6.41 leads to a set of values of e as the roots of the corresponding polynomial, and the lowest value is the best value of the ground state of the system with a basis set of the selected form. The coefficients in the linear combination are then found by solving the set of secular equations with this value of e. The procedure is illustrated in the following example.
Example 6.7
Using the Rayleigh–Ritz method
Suppose we are investigating the effect of mass of the nucleus on the groundstate wavefunctions of the hydrogen atom. One approach might be to use as a trial function a linear combination of the 1s- and 2s-orbitals of a hydrogen atom with an infinitely heavy nucleus but to use the true hamiltonian for the atom. Find the optimum linear combination of these orbitals and the groundstate energy of the atom. Method. We use the wavefunctions of a hydrogen atom with an infinitely
heavy nucleus as the basis, and the hamiltonian of the actual hydrogen atom: neither orbital is an eigenfunction of the hamiltonian, but a linear combination of them can be expected to be a reasonable approximation to an eigenfunction. The first step is to evaluate the matrix elements needed for the secular determinant: these can be expressed in terms of the Rydberg constant R with a suitable correction for the energy. Then set the secular determinant equal to zero and find the lowest root of the resulting polynomial in e. Use this value in the secular equations for the coefficients. Answer. The basis functions are 1=2 1=2 1 1 r r=2a0 r=a0 e c1 ¼ e c ¼ 2 2 a0 pa30 32pa30 The trial function is then ctrial ¼ c1c1 þ c2c2. The basis functions are orthonormal, so S11 ¼ S22 ¼ 1 and S12 ¼ S21 ¼ 0. The hamiltonian is the same as that given in Example 6.6 with Z ¼ 1: H¼
2 2 h e2 r 2m 4pe0 r
and as there, because the basis functions are independent of angles, only the radial derivatives need be retained. Express the energies in terms of hcR ¼ h2 =2a20 me . The integrals required are quite straightforward to evaluate
6.10 THE RAYLEIGH–RITZ METHOD
j
187
and are as follows: H11 ¼ ðg 1ÞhcR H22 ¼ 14ðg 1ÞhcR 16g hcR H12 ¼ H21 ¼ 27 21=2 with g ¼ me/mp. The secular determinant expands as follows: H11 eS11 H12 eS12 H11 e H12 ¼ H eS H22 eS22 H21 H22 e 21 21 ¼ e2 ðH11 þ H22 Þe þ ðH11 H22 H12 H21 Þ ¼0 Substitution of the matrix elements gives the lower root e ¼ 18ðg 1Þf5 þ 3ð1 þ 2G2 Þ1=2 ghcR
where G ¼
26 g 1Þ
34 ðg
Because G ¼ 0.000 43, it follows that e ¼ 0.999 46hcR. The secular equations are c1 ðH11 eÞ þ c2 H21 ¼ 0 c1 H12 þ c2 ðH22 eÞ ¼ 0 and for the trial function to be normalized we also know that c21 þ c22 ¼ 1. It follows that with the value of e found above, c1 1:000 00
c2 ¼ 0:000 54
Comment. The wavefunction has a 3.0 10 5 per cent admixture of
2s-orbital into the 1s-orbital, with a negative sign for the coefficient. The latter signifies a small decrease in amplitude of the overall wavefunction at the nucleus. The explanation of this reduction can be traced to the fact that the reduced mass is slightly less than the mass of the electron, and so the ‘effective particle’ has slightly more freedom than an electron.
The variation principle leads to an upper bound for the energy of the system. It is also possible to use the principle to determine an upper bound for the first excited state by formulating a trial function that is orthogonal to the ground-state function. There are also variational techniques for finding lower bounds, so the true energy can be sandwiched above and below and hence located reasonably precisely. These calculations, though, are often quite difficult because they involve integrals over the square of the hamiltonian. A further remark is that although the variation principle may give a good value for the energy, there is no guarantee that the optimum trial function will give a good value for some other property of the system, such as its dipole moment.
The Hellmann–Feynman theorem Consider a system characterized by a hamiltonian that depends on a parameter P. This parameter might be the internuclear distance in a molecule or the strength of the electric field to which the molecule is exposed. The exact
188
j
6 TECHNIQUES OF APPROXIMATION
(not trial) wavefunction for the system is a solution of the Schro¨dinger equation, so it and its energy also depend on the parameter P. The question we tackle is how the energy of the system varies as the parameter is varied, and we shall now prove the following relation, which is the Hellmann– Feynman theorem: % & dE qH ¼ ð6:42Þ dP qP
Proof 6.2 The Hellmann–Feynman theorem
We suppose that the wavefunction is normalized to 1 for all values of P, in which case Z EðPÞ ¼ cðPÞ HðPÞcðPÞ dt The derivative of E with respect to P is Z Z Z dE qc qH qc ¼ c dt þ c H dt Hc dt þ c dP qP qP qP Z Z Z qc qH qc c dt þ E c dt ¼E c dt þ c qP qP qP Z Z d qH c c dt þ c c dt ¼E dP qP In the last term of the second line, we have employed the hermiticity of H to let it operate on the function standing to its left. The first term on the right of the last line is zero because the integral is equal to 1 for all values of P. The second term is the expectation value of the first-derivative of the hamiltonian.
The great advantage of the Hellmann–Feynman theorem is that the operator qH/qP might be very simple. For example, if the total hamiltonian is H ¼ H(0) þ Px, then qH/qP ¼ x, and there is no mention of H(0), which might be a very complicated operator. In this case, dE ¼ hxi dP and the calculation is apparently very simple. There is, as always, a complication. The proof of the theorem supposes that the wavefunctions are the exact eigenfunctions of the total hamiltonian. Therefore, to evaluate the expectation value of even a simple operator like x, we need to have solved the Schro¨dinger equation for the complete, complicated hamiltonian. Nevertheless, we can use the perturbation theory described earlier in the chapter to arrive at successively better approximations to the true wavefunctions, and therefore can calculate successively better approximations to the value of dE/dP, the response of the system to changes in the hamiltonian. We shall use this technique in Chapters 12 and 13 to calculate the properties of molecules in electric and magnetic fields.
6.11 THE TIME-DEPENDENT BEHAVIOUR OF A TWO-LEVEL SYSTEM
j
189
Time-dependent perturbation theory Just about every perturbation is time-dependent, even those that appear to be stationary. Even stationary perturbations have to be turned on: samples are inserted into electric and magnetic fields, the shapes of vessels are changed, and so on. The reason why time-independent perturbation theory can often be applied in these cases is that the response of a molecule is so rapid that for all practical purposes the systems forget that they were ever unperturbed and settle rapidly into their final perturbed states. Nevertheless, if we really want to understand the properties of molecules, we need to see how systems respond to newly imposed perturbations and then settle into stationary states after an interval. But there is a much more important reason for studying time-dependent perturbations. Many important perturbations never ‘settle down’ to a constant value. A molecule exposed to electromagnetic radiation, for instance, experiences an electromagnetic field that oscillates for as long as the perturbation is imposed. Time-dependent perturbation theory is essential for such problems, and is used to calculate transition probabilities in spectroscopy and the intensities of spectral lines. We adopt the same approach as for time-independent perturbation theory. First, we consider a two-level system. Then we generalize that special case to systems of arbitrary complexity.
6.11 The time-dependent behaviour of a two-level system The total hamiltonian of the system is H ¼ Hð0Þ þ Hð1Þ ðtÞ
ð6:43Þ
A typical example of a time-dependent perturbation is one that oscillates at an angular frequency o, in which case Hð1Þ ðtÞ ¼ 2Hð1Þ cos ot
ð6:44Þ
where H(1) is a time-independent operator and the 2 is present for future convenience. We need to deal with the time-dependent Schro¨dinger equation: HC ¼ ih
qC qt
ð6:45Þ
As in the earlier part of the chapter, we denote the energies of the two states ð0Þ ð0Þ as E1 and E2 and the corresponding time-independent wavefunctions as ð0Þ ð0Þ c1 and c2 . These wavefunctions are the solutions of ð0Þ ð0Þ Hð0Þ cð0Þ n ¼ En cn
ð6:46Þ
and are related to the time-dependent unperturbed wavefunctions by ð0Þ
ð0Þ iEn Cð0Þ n ðtÞ ¼ cn e
t=h
ð6:47Þ
190
j
6 TECHNIQUES OF APPROXIMATION
In the presence of the perturbation H(1)(t), the state of the system is expressed as a linear combination of the basis functions: ð0Þ
ð0Þ
CðtÞ ¼ a1 ðtÞC1 ðtÞ þ a2 ðtÞC2 ðtÞ
ð6:48Þ
Notice that the coefficients are also time-dependent because the composition of the state may evolve with time. The total time-dependence of the wavefunction therefore arises from the oscillation of the basis functions and the evolution of the coefficients. The probability that at any time t the system is in state n is jan(t)j2. Substitution of the linear combination into the Schro¨dinger equation, eqn 6.45, leads to the following expression: ð0Þ
ð0Þ
ð0Þ
ð0Þ
HC ¼ a1 Hð0Þ C1 þ a1 Hð1Þ ðtÞC1 þ a2 Hð0Þ C2 þ a2 Hð1Þ ðtÞC2 q ð0Þ ð0Þ a1 C1 þ a2 C2 ¼ ih qt ð0Þ ð0Þ qC1 qC2 ð0Þ da1 ð0Þ da2 ¼ iha1 þ ihC1 þ iha2 þ ihC2 qt dt qt dt Each basis function satisfies qCð0Þ n qt so the last equation simplifies to Hð0Þ Cð0Þ h n ¼ i
ð0Þ
ð0Þ
ð0Þ
ð0Þ
a1 Hð1Þ ðtÞC1 þ a2 Hð1Þ ðtÞC2 ¼ iha_ 1 C1 þ iha_ 2 C2
where a_ ¼ da/dt. The next step is to extract equations for the time-variation of the coefficients. To do so, we write the time-dependence of the wavefunctions explicitly: ð0Þ
a1 Hð1Þ ðtÞj1ieiE1
t=h
ð0Þ
þ a2 Hð1Þ ðtÞj2ieiE2
ð0Þ iE1 t=h
¼ iha_ 1 j1ie
þ i ha_ 2 j2ie
t=h
ð0Þ iE2 t=h
We have also taken this opportunity to express the wavefunctions cð0Þ n as the kets jni. Now multiply through from the left by h1j and use the orthonormality of the states to obtain ð0Þ
ð1Þ
a1 H11 ðtÞeiE1
t=h
ð1Þ
ð0Þ
þ a2 H12 ðtÞeiE2
t=h
ð0Þ
¼ iha_ 1 eiE1
t=h
ð1Þ
where Hij ðtÞ ¼ hijHð1Þ ðtÞjji. The expression we have obtained can be simplified in a number of ways. ð0Þ ð0Þ In the first place, we shall write ho21 ¼ E2 E1 , and so obtain ð1Þ
ð1Þ
a1 H11 ðtÞ þ a2 H12 ðtÞeio21 t ¼ iha_ 1
ð6:49Þ
Next, it is commonly the case that the time-dependent perturbation has no ð1Þ ð1Þ diagonal elements, so we can set H11 ðtÞ ¼ H22 ðtÞ ¼ 0. The equation then reduces to a_ 1 ¼
1 ð1Þ a2 H12 ðtÞeio21 t ih
ð6:50aÞ
6.11 THE TIME-DEPENDENT BEHAVIOUR OF A TWO-LEVEL SYSTEM
j
191
This differential equation for a1 depends on a2, so we need an equation for that coefficient too. The same procedure, but with multiplication by h2j, leads to a_ 2 ¼
1 ð1Þ a1 H21 ðtÞeio21 t ih
ð6:50bÞ
First, suppose the perturbation is absent, so its matrix elements are zero. In that simple case a_ 1 ¼ 0 and a_ 2 ¼ 0. The coefficients do not change from their initial values and the state is ð0Þ
ð0Þ
CðtÞ ¼ a1 ð0Þc1 eiE1
t=h
ð0Þ
ð0Þ
þ a2 ð0Þc2 eiE2
t=h
ð6:51Þ
Although C(t) oscillates with time, the probability of finding the system in either of the states is constant, because the square modulus of the coefficients of each ai is constant. That is, in the absence of a perturbation, the state of the system is frozen at whatever was its initial composition. Now consider the case of a constant perturbation applied at t ¼ 0 (Fig. 6.9). ð1Þ ð1Þ hV and (by hermiticity) H21 ðtÞ ¼ hV when the We shall write H12 ðtÞ ¼ perturbation is present. Then
(1)
H12 (t )
hV
a_ 1 ¼ iVa2 eio21 t
0
Time, t
T
Fig. 6.9 The form of a constant perturbation switched on at t ¼ 0 and off at t ¼ T.
a_ 2 ¼ iV a1 eio21 t
ð6:52Þ
There are several ways of solving coupled differential equations such as these. The most elementary method (which we employ here) is to substitute one equation into the other.3 On differentiation of a_ 2 and then using the expression for a_ 1 we obtain € a2 ¼ iV a_ 1 eio21 t þ o21 V a1 eio21 t ¼ jVj2 a2 þ io21 a_ 2
ð6:53Þ
The corresponding expression for a¨1 is obtained by differentiating the expression for a_ 1. Note that two coupled first-order equations lead to one second-order differential equation for either a1 or a2. The general solutions of this second-order differential equation are a2 ðtÞ ¼ ðAeiOt þ BeiOt Þeio21 t=2
where O ¼ 12ðo221 þ 4jVj2 Þ1=2
ð6:54Þ
where A and B are constants determined by the initial conditions. A similar expression holds for a1. Now suppose that at t ¼ 0 the system is definitely in state 1. Then a1(0) ¼ 1 and a2(0) ¼ 0. These initial conditions are enough to determine the two constants in the general solution, and after some straightforward algebra we find the following two particular solutions: io21 ijVj sin Ot eio21 t=2 sin Ot eio21 t=2 a1 ðtÞ ¼ cos Ot þ a2 ðtÞ ¼ O 2O ð6:55Þ These are the exact solutions for the problem: we have made no approximations in their derivation.
.......................................................................................................
3. A much more powerful method is to use Laplace transforms.
192
j
6 TECHNIQUES OF APPROXIMATION
6.12 The Rabi formula We are interested in the probability of finding the system in one of the two states as a function of time. These probabilities are P1(t) ¼ ja1(t)j2 and P2(t) ¼ ja2(t)j2. For state 2, the initially unoccupied state, we find the Rabi formula: ! 4jVj2 sin2 12ðo221 þ 4jVj2 Þ1=2 t ð6:56Þ P2 ðtÞ ¼ 2 2 o21 þ 4jVj
Probability, P (t )
1
0
This expression will be at the centre of the following discussion. The probability of the system being in state 1 is of course P1(t) ¼ 1 P2(t), so we do not need to make a special calculation for its value. The first case we consider is that of a degenerate pair of states, so o21 ¼ 0. The probability that the system will be found in state 2 if at t ¼ 0 it was certainly in state 1 is then Time, t
P2 ðtÞ ¼ sin2 jVjt
T
Fig. 6.10 The variation with time of
the probability of being in an initially empty state of a two-level degenerate system that is subjected to a constant perturbation turned on at t ¼ 0 and extinguished at t ¼ T.
Probability, P (t )
1
(a) (b)
0
Time, t
T
Fig. 6.11 The variation with time of
the probability of being in an initially empty state of a two-level non-degenerate system that is subjected to a constant perturbation turned on at t ¼ 0 and extinguished at t ¼ T. The variation labelled (a) corresponds to a small energy separation and that in (b) corresponds to a large separation. Note that the latter oscillates more rapidly than the former.
ð6:57Þ
Figure 6.10 shows a graph of this function. We see that the system oscillates between the two states, and periodically is certainly in state 2. Because the frequency of the oscillation is governed by jVj, we also see that strong perturbations drive the system between its two states more rapidly than weak perturbations. However, provided we wait long enough (specifically, for a time t ¼ p/2jVj), then, whatever the perturbation, in due course the system will be found with certainty in state 2. This responsiveness is a special characteristic of degenerate systems. Degenerate systems are ‘loose’ in the sense that the populations of their states may be transferred completely even by weak stimuli. Now consider the other extreme, when the energy levels are widely separated in comparison with the strength of the perturbation, in the sense o221 4jVj2 . In this case, 4jVj2 can be ignored in both the denominator and the argument of the sine function, and we obtain 2jVj 2 2 1 sin 2o21 t ð6:58Þ P2 ðtÞ o21 The behaviour of the system is now quite different (Fig. 6.11). The populations oscillate, but P2(t) never rises above 4jVj2 /o221 , which is very much less than 1. There is now only a very small probability that the perturbation will drive the system from state 1 to state 2. Moreover, the frequency of oscillation of the population is determined solely by the separation of the states and is independent of the strength of the perturbation. That is like the behaviour of a bell that is struck by a hammer: the frequency is largely independent of the strength of the blow. (Indeed, there is a deep connection between the two phenomena.) The only role of the perturbation, other than its role in causing the transitions, is to govern the maximum extent to which population transfer occurs. If the perturbation is strong (but still weak in comparison with the energy separation of the states), then there is a higher probability of finding the system in state 2 than when the perturbation is weak.
6.13 MANY-LEVEL SYSTEMS: THE VARIATION OF CONSTANTS
Example 6.8
j
193
How to prepare systems in specified states
Suggest how you could prepare a degenerate two-level system in a mixed state in which there is equal likelihood of finding it in either state. Method. We know that a state, once prepared, persists with constant
composition in the absence of a perturbation. This suggests that we should use the Rabi formula to find the time for which a perturbation should be applied to result in P2(t) ¼ 0.5, and then immediately extinguish the perturbation. Answer. The Rabi formula shows that P2(t) ¼ 0.5 when t ¼ p/4jVj. Therefore, the perturbation should be applied to a system that is known to be in state 1 initially, and removed at t ¼ p/4jVj. Although the wavefunction of the system will oscillate, the probability of finding the system in either state will remain 0.5 until another perturbation is applied. Comment. This state preparation procedure is the quantum mechanical basis
of pulse techniques in nuclear magnetic resonance. Self-test 6.8. For how long should the perturbation be applied to the
same system to obtain a state with probability 0.25 of being in state 2?
6.13 Many-level systems: the variation of constants The discussion of the two-level system has revealed two rather depressing features. One is that even very simple systems lead to very complicated differential equations. For a two-level system the problem requires the solution of a second-order differential equation; for an n-level system, the solution requires dealing with an nth-order differential equation, which is largely hopeless. The second point is that even for a two-level system, the differential equation could be solved only for a trivially simple perturbation, one that did not vary with time. The differential equation is very much more complicated to solve when the perturbation has a realistic time-dependence, such as oscillation in time. Even the case cos ot is very complicated. Clearly, we need to set up an approximation technique for dealing with systems of many levels and which can cope with realistic perturbations. We shall describe the technique invented by P.A.M. Dirac and known (agreeably paradoxically) as the variation of constants. It is a generalization of the two-level problem, and that relationship should be held in mind as we go through the material. As before, the hamiltonian is taken to be H ¼ H(0) þ H(1)(t). The eigenstates of H(0) will be denoted by the ket jni or by the corresponding wavefunction cð0Þ n as convenient, where ð0Þ
ð0Þ iEn Cð0Þ n ðtÞ ¼ cn e
t=h
Hð0Þ Cð0Þ h n ¼ i
qCð0Þ n qt
194
j
6 TECHNIQUES OF APPROXIMATION
The state of the perturbed system is C. As before, we express it as a timedependent linear combination of the time-dependent unperturbed states: CðtÞ ¼
X
an ðtÞCð0Þ n ðtÞ ¼
n
X
ð0Þ
iEn an ðtÞcð0Þ n e
t=h
HC ¼ ih
n
qC qt
ð6:59Þ
Our problem, as for the two-level case, is to find how the linear combination evolves with time. To do so, we set up and then solve the differential equations satisfied by the coefficients an. We proceed as before. Substitution of C into the Schro¨dinger equation leads to the following expressions: X X an ðtÞ Hð0Þ Cð0Þ an ðtÞHð1Þ ðtÞCð0Þ HC ¼ n ðtÞ þ n ðtÞ |fflfflfflfflfflfflffl ffl {zfflfflfflfflfflfflffl ffl } n n #
ih
qC ¼ qt
X n
an ðtÞih
X qCð0Þ n þ ih a_ n ðtÞCð0Þ n ðtÞ qt n
The two indicated terms are equal, so we are left with X X a_ n ðtÞCð0Þ an ðtÞHð1Þ ðtÞCð0Þ h n ðtÞ¼ i n ðtÞ n
n
In terms of the time-independent kets, this equation is X X ð0Þ ð0Þ a_ n ðtÞjnieiEn t=h an ðtÞHð1Þ ðtÞjnieiEn t=h ¼ ih n
n
At this point we have to extract one of the a_ n on the right. To do so, we make use of the orthonormality of the eigenstates, and multiply through by hkj: X ð0Þ ð0Þ an ðtÞhkjH ð1Þ ðtÞjnieiEn t=h ¼ i ha_ k ðtÞeiEk t=h n ð1Þ
We simplify the appearance of this expression by writing Hkn ðtÞ ¼ ð0Þ ð0Þ hokn ¼ Ek En , when it becomes hkjHð1Þ ðtÞjni and a_ k ðtÞ ¼
1X ð1Þ an ðtÞHkn ðtÞeiokn t ih n
ð6:60Þ
Equation 6.60 is exact. We can move towards finding exact solutions and from this point on the development diverges from the exact two-level calculation described earlier. To solve a first-order differential equation, we integrate it from t ¼ 0, when the coefficients had the values an(0), to the time t of interest: Z 1X t ð1Þ ak ðtÞ ak ð0Þ ¼ an ðtÞHkn ðtÞeiokn t dt ð6:61Þ ih n 0 The trouble with this equation is that although it appears to give an expression for any coefficient ak(t), it does so in terms of all the coefficients, including ak itself. These other coefficients are unknown, and must be determined from equations of a similar form. So, to solve eqn 6.61, it appears
6.14 THE EFFECT OF A SLOWLY SWITCHED CONSTANT PERTURBATION
f
(a)
(b)
(c)
i
the text corresponds to considering only direct transitions between the initial and final states (as in (a)), and ignoring indirect transitions (as in (b) and (c)), which correspond to higherorder processes.
H
(1)
(a) i
f
H (1)
H (1)
(b) n
i
f
Fig. 6.13 Diagrams for (a) first-order and (b) second-order contributions to the perturbation of a system. (1)
Perturbation, H (1)(t )
H
195
that we must already know all the coefficients! A way out of this cyclic problem is to make an approximation. We shall base the approximation on the supposition that the perturbation is so weak and applied for so short a time that all the coefficients remain close to their initial values. Then, if the system is certainly in state jii at t ¼ 0, all coefficients other than ai are close to zero throughout the period for which the perturbation is applied, and any single coefficient, such as the coefficient of state jfi that is zero initially, is given by af ðtÞ ¼
Fig. 6.12 The procedure described in
j
1 ih
Z
t
0
ð1Þ
ai ðtÞHfi ðtÞeiofi t dt
because all terms in the sum are zero (an(t) 0) except for the term corresponding to the initial state. We have also made use in the sum of the fact that af (t) af(0) ¼ 0. However, the coefficient of the initial state remains close to 1 for all the time of interest, so we can set ai(t) 1, and obtain af ðtÞ ¼
1 ih
Z 0
t
ð1Þ
Hfi ðtÞeiofi t dt
ð6:62Þ
This is an explicit expression for the value of the coefficient of a state that was initially unoccupied and will be the formula that we employ in the following discussion. The approximation we have adopted ignores the possibility that the perturbation can take the system from its initial state jii to some final state jfi by an indirect route in which the perturbation induces a sequence of several transitions (Fig. 6.12). Put another way: the approximation assumes that the perturbation acts only once, and that we are therefore dealing with first-order perturbation theory. This restriction to first-order contributions can be expressed diagrammatically (Fig. 6.13): the intersection of the sloping and horizontal lines is intended to convey the idea that the perturbation (the sloping line) acts on the molecular states (the horizontal line) only once. The upper diagram in Fig. 6.13 can be regarded as a succinct expression for the right-hand side of eqn 6.62. Second-order perturbation theory (which we are not doing here) would give rise to diagrams like the one shown in the lower part of Fig. 6.13. These diagrams are sometimes associated with the name of R.P. Feynman, who introduced similar diagrams in the context of fundamental particle interactions, and are called Feynman diagrams.
6.14 The effect of a slowly switched constant perturbation
0
Time, t
Fig. 6.14 An exponentially switched
but otherwise constant perturbation.
As a first example of how to use eqn 6.62, consider a perturbation that rises slowly from zero to a steady final value (Fig. 6.14). Such a switched perturbation is H
ð1Þ
ðtÞ ¼
0 Hð1Þ ð1 ekt Þ
for t < 0 for t 0
ð6:63Þ
196
j
6 TECHNIQUES OF APPROXIMATION
where H(1) is a time-independent operator and, for slow switching, k is small (and positive). The coefficient of an initially unoccupied state is given by eqn 6.62 as Z 1 ð1Þ t af ðtÞ ¼ Hfi 1 ekt eiofi t dt ih 0 ð6:64Þ 1 ð1Þ eiofi t 1 eðkiofi Þt 1 þ ¼ Hfi ih iofi k iofi This result, which is exact within first-order perturbation theory, can be simplified by supposing that we are interested in times very long after the perturbation has reached its final value, which means t 1/k, and that the perturbation is switched slowly in the sense that k2 o2fi . Then ð1Þ 2 Hfi ð6:65Þ jaf ðtÞj2 ¼ 2 h o2fi This is the result that would have been obtained by applying timeindependent perturbation theory (compare to eqn 6.21), and assuming that the constant perturbation had always been present. We can now see why time-independent perturbation theory can be used for most problems of chemical interest, except where the perturbation continues to change after it has been applied. When a ‘constant’ perturbation is switched on, it is done so very slowly in comparison with the frequencies associated with the transitions in atoms and molecules (k 103 s1, of i 1015 s1). Furthermore, we are normally interested in a system’s properties at times long after the switching is complete (t 10 3 s; and in general kt 1). These are the conditions under which time-dependent perturbation theory has effectively settled down into time-independent perturbation theory. All the transients stimulated by the switching have subsided and the populations of states are steady. Example 6.9
The effect of a constant perturbation
A constant perturbation was switched on exponentially starting at t ¼ 0. Evaluate the probability of finding a system in state 2 given that initially it was in state 1, and illustrate the role of transients. Method. The perturbation is given by eqn 6.63 and the solution is expressed
by eqn 6.64. To find the probability that the system is in state 2, we need to form P2 ¼ ja2(t)j2 for a general value of k and then to plot P2 against t. ð1Þ For example plots, set l ¼ k/o21 and plot P2/(jVj/o21)2, with jVj ¼ H12 / h, for l ¼ 0.01, 0.1, and 1, which corresponds to switching rates increasing in 10-fold steps. Answer. From eqn 6.64 with l ¼ k/o21 and x ¼ o21t, P2 ðtÞ ¼
jV j2 p2 ðtÞ o221
6.15 THE EFFECT OF AN OSCILLATING PERTURBATION
197
with
3 Probability, P (t )/(|V |2/ 21)2
j
' (* 1 ) 1 þ 2l2 2l2 cos x þ 2 elx elx þ 2l 1 elx sin x 2 1þl This function is plotted for l ¼ 0.01, 0.1, 1 in Fig. 6.15. p2 ðtÞ ¼
2
=1
Comment. Notice how slow switching (l ¼ 0.01) generates hardly any tran-
sients, whereas rapid switching (l ¼ 1) is like an impulsive shock to the system, and causes the population to oscillate violently between the two states. For very rapid switching (l 1), p2 varies as 2(1 cos x), and so it oscillates between 0 and 4 with an average value of 2: such rapid switching is like a hammer blow.
= 0.1
1
= 0.01
0
0
20
40
60 21t
80
Self-test 6.9. Suppose the constant perturbation was switched on as l hVt for 100
0 lt < 1 and remained at hV for lt 1. Investigate how the transients behave.
Fig. 6.15 The time variation of the
probability of occupying an initially0unoccupied state when the0perturbation is switched on at different rates for different values of0the switching rate as expressed by the parameter l ¼ k/o21.
6.15 The effect of an oscillating perturbation We now consider a system that is exposed to an oscillating perturbation, such as an atom may experience when it is exposed to electromagnetic radiation in a spectrometer or in sunlight. Once we can deal with oscillating perturbations, we can deal with all perturbations, for a general time-dependent perturbation can be expressed as a superposition of harmonically oscillating functions. In the first stage of the discussion we consider transitions between discrete states jii and jfi. A perturbation oscillating with an angular frequency o ¼ 2pn and turned on at t ¼ 0 has the form ð6:66Þ Hð1Þ ðtÞ ¼ 2Hð1Þ cos ot ¼ Hð1Þ eiot þ eiot for t 0. If this perturbation is inserted into eqn 6.62 we obtain Z 1 ð1Þ t iot af ðtÞ ¼ Hfi e þ eiot eiofi t dt ih 0 ð6:67Þ 1 ð1Þ eiðofi þoÞt 1 eiðofi oÞt 1 þ ¼ Hfi ih iðofi þ oÞ iðofi oÞ As it stands, eqn 6.67 is quite obscure (but it is quite easy to compute). It can be simplified to bring out its principal content by taking note of the conditions under which it is normally used. In applications in electronic spectroscopy, the frequencies of i and o are of the order of 1015 s1; in NMR, the lowest frequency form of spectroscopy generally encountered, the frequencies are still higher than 106 s1. The exponential functions in the numerator of the term in braces are of the order of 1 regardless of the frequencies in its argument (because eix ¼ cos x þ i sin x, and neither harmonic function can exceed 1). However, the denominator in the first term is of the order of the frequencies, so the first term is unlikely to be larger than about 106 and may be of the order of 1015 in electronic spectroscopy. In contrast, the denominator in
198
j
6 TECHNIQUES OF APPROXIMATION
the second term can come arbitrarily close to 0 as the external perturbation approaches a transition frequency of the system. Therefore, the second term is normally larger than the first for absorption, and overwhelms it completely as the frequencies approach one another. Consequently, in most practical applications we can be confident about ignoring the first term. When that is done, it is easy to conclude that the probability of finding the system in the discrete state jfi after a time t if initially it was in state jii at t ¼ 0 is ð1Þ 2 4Hfi sin2 12ðofi oÞt ð6:68Þ Pf ðtÞ ¼ 2 h ðofi oÞ2
h
i
f
ð1Þ
Once again we write jHfi j2 ¼ h2 jVfi j2 , in which case we obtain
E i = h ( i +) E f = h f
Pf ðtÞ ¼
Fig. 6.16 The use of an oscillating
perturbation effectively modifies the energy separation between the initial and final states, and at resonance the overall system is effectively degenerate and hence highly responsive.
4jVfi j2 ðofi oÞ2
ð0Þ
Transition probability
Time4
Time2 Time1
0
fi –
Fig. 6.17 The variation of
transition probability with offset frequency and time. Note that the central portion of the curve becomes taller but narrower with time.
ð6:69Þ
The last expression should be familiar. Apart from a small but significant modification, it is exactly the same as eqn 6.58, the expression for a static perturbation applied to a two-level system. The one significant difference is that instead of the actual frequency difference of i appearing in the expression, it is replaced throughout by ofi o. This replacement can be interpreted as an effective shift in the energy differences involved in exciting the system as a result of the presence of a photon in the electromagnetic field. As depicted in Fig. 6.16, where the wavy line now represents an oscillating perturbation, ð0Þ ð0Þ the overall energy difference Ef Ei should actually be thought of as ð0Þ
Ef Ei
Time3
sin2 12ðofi oÞt
¼ E(excited molecule, no photon) Eðground-state molecule, photon of energy hoÞ ¼ hðofi oÞ
According to eqn 6.69, the time-dependence of the probability of being found in state jfi depends on the frequency offset, ofi o (Fig. 6.17). When the frequency offset is zero, the field and the system are said to be in resonance, and the transition probability increases most rapidly with time. To obtain the quantitative form of the time-dependence at resonance, we take the limit of eqn 6.69 as o ! ofi by using x 16x3 þ sin x ¼1 ¼ lim x!0 x x!0 x lim
Then, lim Pf ðtÞ ¼ jVfi j2 t2
o!ofi
ð6:70Þ
and the probability increases quadratically with time. This conclusion is valid so long as jVfij2t2 1, because that is the underlying assumption of first-order perturbation theory. It follows that the transition probability may approach (and, indeed, in this approximation, unphysically exceed) 1 as the applied frequency approaches a transition frequency. This behaviour can be interpreted in terms of the system then becoming, in effect, a loose, degenerate
6.16 TRANSITION RATES TO CONTINUUM STATES ð0Þ
j
199
ð0Þ
system as the overall energy difference Ef Ei approaches zero, and which can be nudged fully from state to state even by gentle perturbations.
6.16 Transition rates to continuum states We now turn to the case in which the final state is a part of a continuum of states. Although we can still use eqn 6.69 to calculate the transition probability to one member of the continuum, the observed transition rate is an integral over all the transition probabilities to which the perturbation can drive the system. Specifically, if the density of states is written r(E), where r(E)dE is the number of continuum states in the range E to E þ dE, then the total transition probability, P(t), is Z Pf ðtÞrðEÞ dE ð6:71Þ PðtÞ ¼ range
In this expression ‘range’ means that the integration is over all final states accessible under the influence of the perturbation. To evaluate the integral, we first express the transition frequency of i in terms of the energy E by writing of i ¼ E/h Z sin2 12ðE= h oÞt 4jVfi j2 rðEÞ dE PðtÞ ¼ ðE= h oÞ2 range
Transition probability
The integral can be simplified by noting that the factor (sin2x)/x2 is sharply peaked close to E/h ¼ o, the frequency of the radiation. However, for an appreciable transition probability, the frequency of the incident radiation must be close to the transition frequency ofi, so we can set E/h ofi wherever hofi, E occurs. In other words, we can evaluate the density of states at Efi ¼ and treat it as a constant. Moreover, although the matrix elements jVfij vary with E, such a narrow range of energies contributes to the integral that it is permissible to treat jVfij as a constant. The integral then simplifies to Z 4 sin2 12ðE= h oÞt PðtÞ ¼ jVfi j2 rðEfi Þ dE ðE= h oÞ2 range
0
fi –
Fig. 6.18 The extension of the
range of integration from the actual range (light shading) to infinity (dark shading) barely affects the value of the integral.
An additional approximation that stems from the narrowness of the function remaining in the integrand is to extend the limits from the actual range to infinity: the integrand is so small outside the actual range that this extension introduces no significant error (Fig. 6.18). At this point it is also convenient h o)t, which implies that dE ¼ (2 h/t)dx. Consequently, the to set x ¼ 12(E/ integral becomes Z 1 2 2 h sin x dx PðtÞ ¼ jVfi j2 rðEfi Þt2 2 t 1 x The integral is standard: Z 1 2 sin x dx ¼ p 2 1 x
200
j
6 TECHNIQUES OF APPROXIMATION
Therefore, we conclude that PðtÞ ¼ 2p htjVfi j2 rðEfi Þ
ð6:72Þ
which increases linearly with time. The physical reason for this different timedependence (compared to the result in eqn 6.70) is that as time increases, the height of the central peak in Fig. 6.17 increases as t2, but the width of the central peak decreases, and is proportional to 1/t. The area under the curve therefore increases as t2 1/t ¼ t. The transition rate, W, is the rate of change of probability of being in an initially empty state: W¼
dP dt
ð6:73Þ
and the intensities of spectral lines are proportional to these transition rates because they depend on the rate of transfer of energy between the system and the electromagnetic field. It follows that W ¼ 2p hjVfi j2 rðEfi Þ
ð6:74Þ
This succinct expression is called Fermi’s golden rule. It asserts that to calculate a transition rate, all we need do is to multiply the square modulus of the transition matrix element between the two states by the density of states at the transition frequency.
6.17 The Einstein transition probabilities Einstein considered the problem of the transfer of energy between the electromagnetic field and matter and arrived at the conclusion that although eqn 6.74 correctly accounts for the absorption of radiation, it fails to take into account all contributions to the emission of radiation from an excited state. He considered a collection of atoms that were in thermal equilibrium with the electromagnetic field at a temperature T. First, we note that the quantity jVf ij2 is proportional to the square of the electric field strength of the incident radiation (for a perturbation of the form –me), and hence is proportional to the intensity, I, of the radiation at the frequency of the transition. The intensity is defined so that the energy of radiation in the frequency range n to n þ dn that passes through an area A in an interval Dt is
A c∆t
dE ¼ IðnÞADt dn
ð6:75Þ
Because all the radiation within a distance cDt can pass through the area A in that time interval (Fig. 6.19), the volume containing the energy is cDtA, and the energy density, rrad(n)dn, in that frequency range is Fig. 6.19 All photons within a
distance cDt can reach the right-hand wall in an interval Dt.
rrad ðnÞdn ¼
dE IðnÞ ¼ dn AcDt c
6.17 THE EINSTEIN TRANSITION PROBABILITIES
j
201
Therefore, the energy density of radiation, the energy in a given volume and given frequency range divided by the volume of the region and the range of frequencies, is rrad ðnÞ ¼
IðnÞ c
ð6:76Þ
Consequently, jVfij2 is proportional to rrad evaluated at the transition frequency, or equivalently through the relation Ef i ¼ hnfi, at the transition energy. It follows that we can write Wf
i
¼ Bif rrad ðEfi Þ
ð6:77Þ
where Bif is the Einstein coefficient of stimulated absorption. Einstein also recognized that the rate at which an excited state jfi is induced to make transitions down to the ground state jii is also proportional to the intensity of radiation at the transition frequency: Wf!i ¼ Bfi rrad ðEfi Þ
ð6:78Þ
The coefficient Bfi is the Einstein coefficient of stimulated emission. It is simple to show that Bif ¼ Bfi. The argument is based on the hermiticity of the perturbation hamiltonian, which lets us write Bif / Vif Vif ¼ Vfi Vfi / Bfi Einstein, however, was able to infer this equality in a different way, as we shall now see. Specifically, for electric-dipole allowed transitions, we show in Further information 16 that Bif ¼
jmfi j2 6e0 h2
where fi is the transition dipole moment: Z fi ¼ cf ci dt
ð6:79Þ
ð6:80Þ
with the electric dipole moment operator. The transition probabilities we have derived refer to individual atoms. If there are Ni atoms in the state jii and Nf in the state jfi, then at thermal equilibrium, when there is no net transfer of energy between the system and the field, Ni Wf
i
¼ Nf Wf!i
Because the two transition rates are equal, it follows that the populations are also equal. However, that conclusion is in conflict with the Boltzmann distribution, which requires from very general principles that Nf ¼ eEfi =kT Ni
202
j
6 TECHNIQUES OF APPROXIMATION
To avoid this conflict, Einstein proposed that there was an additional contribution to the emission process that is independent of the presence of radiation of the transition frequency. This additional contribution he wrote spont ¼ Afi Wf!i
ð6:81Þ
where Afi is the Einstein coefficient of spontaneous emission. The total rate of emission is therefore Wf!i ¼ Afi þ Bfi rrad ðEfi Þ
ð6:82Þ
and the condition for thermal equilibrium is now Ni Bif rrad ðEfi Þ ¼ Nf fAfi þ Bfi rrad ðEfi Þg This expression is consistent with the Boltzmann distribution. Indeed, if we accept the Boltzmann distribution for the ratio Nf /Ni, it can be rearranged into rrad ðEfi Þ ¼
Afi =Bfi ðBif =Bfi ÞeEfi =kT 1
However, it is also known from very general considerations that at equilibrium, the density of states of the electromagnetic field is given by the Planck distribution (see the Introduction): rrad ðEfi Þ ¼
8phn3fi =c3 eEfi =kT 1
ð6:83Þ
Comparison of the last two expressions confirms that Bif ¼ Bfi and, moreover, gives a relation between the coefficients of stimulated and spontaneous emission: Afi ¼
8phn3fi Bfi c3
ð6:84Þ
The important point about eqn 6.84 is that it shows that the relative importance of spontaneous emission increases as the cube of the transition frequency, and that it is therefore potentially of great importance at very high frequencies. That is one reason why X-ray lasers are so difficult to make: highly excited populations are difficult to maintain and discard their energy at random instead of cooperating in a stimulated emission process. The spontaneous emission process can be viewed as the outcome of the presence of zero-point fluctuations of the electromagnetic field. As indicated in footnote 2 of Chapter 7 (Section 7.3), the electromagnetic field has zero-point oscillations even though there are no photons present. These fluctuations perturb the excited state and induce the transition to
6.18 LIFETIME AND ENERGY UNCERTAINTY
j
203
Amplitude, re Ψ
a lower state. ‘Spontaneous’ transitions are actually caused by these zero-point fluctuations of the electromagnetic vacuum. Spontaneous absorptions in a field devoid of photons are ruled out by the conservation of energy.
t
Fig. 6.20 A wavefunction
corresponding to a precise energy has a constant maximum amplitude; if the wavefunction decays, then it no longer corresponds to a precise energy.
The expression for g given in eqn 6.87 can be verified by using Euler’s relation and the definite integrals Z 1 sin½aðb xÞ dx x 2 þ c2 1 p ¼ eac sin ab c Z 1 cos½aðb xÞ dx x2 þ c2 1 p ¼ eac cos ab c A function of the form a f ðxÞ ¼ 2 x þ b2 0.1
6.18 Lifetime and energy uncertainty We are now in a position to establish the relation between the lifetime of a state and the range of energies that it may possess. We have seen that if a state has a precise energy, then its time-dependent wavefunction has the form C ¼ ce iEt/h; such states are stationary states in the sense that jCj2 ¼ jcj2, a time-independent probability density. However, if the wavefunction decays with time, perhaps because the system is making transitions to other states, then its energy is imprecise. We suppose that the probability of finding the system in a particular excited state decays exponentially with time with a time-constant t: jCj2 ¼ jcj2 et=t
ð6:85Þ
The justification of this assumption can be found in the references in Further reading and Section 14.13. The amplitude therefore has the form C ¼ ceiEt=ht=2t
ð6:86Þ
This wavefunction decays as it oscillates (Fig. 6.20), and its energy is not immediately obvious. However, such a function can be modelled as a superposition of oscillating functions by using the techniques of Fourier analysis, and we write Z 0 eiEt=ht=2t ¼ gðE0 ÞeiE t=h dE0 where gðE0 Þ ¼
ð h=2ptÞ ðE E0 Þ2 þ ð h=2tÞ2
ð6:87Þ
a = 10, b = 10 f (x) a = 20, b = 20
0.05
0
–40 –20
0 x
20
40
has a maximum value f(x ¼ 0) ¼ a/ b2 and has its half-height a/2b2 at x ¼ b. The illustration shows a graph of the function for two sets of values of a and b.
This expression shows that the decaying function corresponds to a range of energies (in fact, all values of energy appear in the superposition), and therefore it implies that any state that has a finite lifetime must be regarded as having an imprecise energy. We can arrive at the quantitative relation between lifetime and energy by considering the shape of the spectral density function, g (Fig. 6.21). The width at half-height is readily shown to be equal to h/2t, and this quantity can be taken as an indication of the range of energies dE present in the state. It follows that h tdE 12
ð6:88Þ
204
j
6 TECHNIQUES OF APPROXIMATION
This lifetime broadening relation is reminiscent of the uncertainty principle (Sections 1.16 and 1.18). It shows that the shorter the lifetime of the state (the shorter the time-constant t for its decay), then the less precise its energy. When a state has zero lifetime, we can say nothing about its energy. Only when the lifetime of a state is infinite can the energy be specified exactly.
Spectral density, g (x )
2
0.5 1
1.0 0
0
1
x
2
3
Fig. 6.21 The spectral density
function for two wavefunctions that decay at different rates. The labels of the lines are the values of h/2t, with x ¼ E E 0 (in the same units).
PROBLEMS 6.1 One excited state of the sodium atom lies at 25 739.86 cm1 above the ground state, another lies at 50 266.88 cm1. Suppose they are connected by a perturbation equivalent in energy to (a) 100 cm1, (b) 1000 cm1, (c) 5000 cm1. Calculate the energies and composition of the states of the perturbed system. Hint. Use eqn 6.6 for the energies and eqn 6.8 for the states, and express the composition as the contribution of the unperturbed states. 6.2 A simple calculation of the energy of the helium atom supposes that each electron occupies the same hydrogenic 1s-orbital (but with Z ¼ 2). The electron–electron interaction is regarded as a perturbation, and calculation gives
Z
c21s ðr1 Þ
2 e2 5 e c21s ðr2 Þdt ¼ 4 4pe0 a0 4pe0 r12
(see Example 7.3). Estimate (a) the binding energy of helium, (b) its first ionization energy. Hint. Use eqn 6.6 with E1 ¼ E2 ¼ E1s. Be careful not to count the electron–electron interaction energy twice. 6.3 Show that the energy of the perturbed levels is related ¼ 1/2 to the mean energy of the unperturbed levels E ¼ 1(E1 E2) sec 2z, where z is the (E1 þ E2) by E E 2 parameter in eqn 6.9. Devise a diagrammatic method of depends on E1 E2 and z. Hint. Use showing how E E eqn 6.9.
6.4 We normally think of the one-dimensional well as being horizontal. Suppose it is vertical; then the potential energy of the particle depends on x because of the presence of the gravitational field. Calculate the first-order correction to the zero-point energy, and evaluate it for an electron in a box on the surface of the Earth. Account for the result. Hint. The energy of the particle depends on its height as mgx where g ¼ 9.81 m s2. Use eqn 6.20 with c(x) given by n ¼ 1 in eqn 2.31. Because g is so small, the energy correction is tiny; but it would be significant if the box were on the surface of a neutron star. 6.5 Calculate the second-order correction to the energy for the system described in Problem 6.4 and calculate the ground-state wavefunction. Account for the shape of the distortion caused by the perturbation. Hint. Use eqn 6.24 for the energy and eqn 6.22 for the wavefunction. The integrals involved are of the form
Z Z
x sin ax sin bx dx ¼
cos ax sin bx dx ¼
d da
Z
cos ax sin bx dx
cosða bÞx cosða þ bÞx 2ða bÞ 2ða þ bÞ
Evaluate the sum over n numerically. 6.6 Calculate the first-order correction to the energy of a ground-state harmonic oscillator subject to an anharmonic potential of the form ax3 þ bx4 where a and b are small
PROBLEMS
j
205
(anharmonicity) constants. Consider the three cases in which the anharmonic perturbation is present (a) during bond expansion (x 0) and compression (x 0), (b) during expansion only, (c) during compression only.
functions of the form (a) sin kx, (b) (x x2/L) þ 1 1 k(x x2/L)2, (c) e k(x 2L) e 2kL for x 12L, and k(x 12L) 12kL 1 e e for x 2L. Find the optimum values of k and the corresponding energies.
6.7 In the free-electron molecular orbital method (Problem 2.19) the potential energy may be made slightly more realistic by supposing that it varies sinusoidally along the polyene chain. Select a potential energy with suitable periodicity, and calculate the firstorder correction to the wavelength of the lowest energy transition.
6.14 Consider the hypothetical linear H3 molecule. The wavefunctions may be modelled by expressing them as c ¼ cAsA þ cBsB þ cCsC, the si denoting hydrogen 1s-orbitals of the relevant atom. Use the Rayleigh–Ritz method to find the optimum values of the coefficients and the energies of the orbital. Make the approximations Hss ¼ a, Hss 0 ¼ b for neighbours but 0 for non-neighbours, Sss ¼ 1, and Sss 0 ¼ 0. Hint. Although the basis can be used as it stands, it leads to a 3 3 determinant and hence to a cubic equation for the energies. A better procedure is to set up symmetry-adapted combinations, and then to use the vanishing of Hij unless Gi ¼ Gj.
6.8 Show group-theoretically that when a perturbation of the form H(1) ¼ az is applied to a hydrogen atom, the 1s-orbital is contaminated by the admixture of npz-orbitals. Deduce which orbitals mix into (a) 2px-orbitals, (b) 2pz-orbitals, (c) 3dxy-orbitals. 6.9 The symmetry of the ground electronic state of the water molecule is A1. (a) An electric field, (b) a magnetic field is applied perpendicular to the molecular plane. What symmetry species of excited states may be mixed into the ground state by the perturbations? Hint. The electric interaction has the form H(1) ¼ ax; the magnetic interaction has the form H(1) ¼ blx. 6.10 Repeat Problem 6.5, but estimate the second-order energy correction using the closure approximation. Compare the two calculations and deduce the appropriate value of DE. Hint. Use eqn 6.27. 6.11 Calculate the second-order energy correction to the ground state of a particle in a one-dimensional box for a perturbation of the form H(1) ¼ e sin(px/L) by using the closure approximation. Infer a value of DE by comparison with the numerical calculation in the Example 6.2. These two problems (6.10 and 6.11) show that the parameter DE depends on the perturbation and is not simply a characteristic of the system itself. 6.12 Suppose that the potential energy of a particle on a ring depends on the angle f as H(1) ¼ e sin2 f. Calculate the first-order corrections to the energy of the degenerate ml ¼ 1 states, and find the correct linear combinations for the perturbation calculation. Find the second-order correction to the energy. Hint. This is an example of degenerate-state perturbation theory, and so find the correct linear combinations by solving eqn 6.35 after deducing the energies from the roots of the secular determinant. For the matrix elements, express sin f as (1/2i)(eif eif). When evaluating eqn 6.35, do not forget the ml ¼ 0 state lying beneath the degenerate pair. The energies are equal to ml2 h2/2mr2; use cml ¼ 1=2 iml f ð1/2pÞ e for the unperturbed states. 6.13 A particle of mass m is confined to a one-dimensional square well of the type treated in Chapter 2. Choose trial
6.15 Repeat the last problem but set HsA sC ¼ g and Sss 0 6¼ 0. Evaluate the overlap integrals between 1s-orbitals on centres separated by R; use
(
S¼
) R 1 R 2 R=a0 e 1þ þ a0 3 a0
Suppose that b=g ¼ SsA sB =SsA sC . For a numerical result, take R ¼ 80 pm, a0 ¼ 53 pm. 6.16 A hydrogen atom in a 2s1 configuration passes into a region where it experiences an electric field in the z-direction for a time t. What is its electric dipole moment during its exposure and after it emerges? Hint. Use eqn 6.55 with o21 ¼ 0; theR dipole moment is the expectation value of ez; use c2s zc2pz dt ¼ 3a0 . 6.17 A biradical is prepared with its two electrons in a singlet state. A magnetic field is present, and because the two electrons are in different environments their interaction with the field is (mB/ h)B(gls1z þ g2s2z) with gl 6¼ g2. Evaluate the time-dependence of the probability that the electron spins will acquire a triplet configuration (that is, the probability that the S ¼ 1, MS ¼ 0 state will be populated). Examine the role of the energy separation hJ of the singlet state and the MS ¼ 0 state of the triplet. Suppose g1 g2 1 103 and J 0; how long does it take for the triplet state to emerge when B ¼ 1.0 T? Hint. Use eqn 6.56; take j0,0i ¼ (1/21/2)(ab ba) and j1,0i ¼ (1/21/2)(ab þ ba). See Problem 4.24 for the significance of mB and g. 6.18 An electric field in the z-direction is increased linearly from zero. What is the probability that a hydrogen atom, initially in the ground state, will be found with its electron in a 2pz-orbital at a time t? Hint. Use eqn 6.62 ð1Þ with Hfi / t. 6.19 At t ¼ 12T the strength of the field used in Problem 6.18 begins to decrease linearly. What is the probability that the electron is in the 2pz-orbital at t ¼ T?
206
j
6 TECHNIQUES OF APPROXIMATION
What would the probability be if initially the electron was in a 2s-orbital? 6.20 Instead of the perturbation being switched linearly, it was switched on and off exponentially and slowly, the switching off commencing long after the switching on was complete. Calculate the probabilities, long after the perturbation has been extinguished, of the 2pz-orbital being occupied, the initial states being as in Problem 6.18. Hint. Take H(1) / 1 ekt for 0 t T and H(1) / ek(t T) for t T. Interpret ‘slow’ as k o and ‘long after’ as both kT 1 (for ‘long after switching on’) and k(t T) 1 (for ‘long after switching off’). 6.21 Calculate the rates of stimulated and spontaneous emission for the 3p ! 2s transition in hydrogen when it is inside a cavity at 1000 K.
6.22 Find the complete dependence of the A and B coefficients on atomic number for the 2p ! 1s transitions of hydrogenic atoms. Calculate how the stimulated emission rate depends on Z when the atom is exposed to black-body radiation at 1000 K. Hint. The relevant density of states also depends on Z. 6.23 Examine how the A and B coefficients depend on the length of a one-dimensional square well for the transition n þ 1 ! n. 6.24 Estimate the lifetime of the upper state of a spectroscopic transition if the spectra shows a peak with a full width at half maximum of (a) 0.010 cm1, (b) 1.5 cm1, (c) 40 cm1. Hint. Use eqn 6.88.
7
Atomic spectra and atomic structure
The spectrum of atomic hydrogen 7.1 The energies of the transitions 7.2 Selection rules 7.3 Orbital and spin magnetic moments 7.4 Spin–orbit coupling 7.5 The fine-structure of spectra 7.6 Term symbols and spectral details 7.7 The detailed spectrum of hydrogen The structure of helium 7.8 The helium atom 7.9 Excited states of helium 7.10 The spectrum of helium 7.11 The Pauli principle
A great deal of chemically interesting information can be obtained by interpreting the line spectra of atoms, the frequencies of the electromagnetic radiation that atoms emit when they are excited. We can use the information to establish the electronic structures of the atoms, and then use that information as a basis for discussing the periodicity of the elements and the structures of the bonds they form. Atomic spectra were also of considerable historical importance, because their study led to the formulation of the Pauli principle, without which it would be impossible to understand atomic structure, chemical periodicity, and molecular structure. The information provided by atoms is of considerable importance for the discussion of molecular structure. For example, we need values of ionization energies and spin–orbit coupling parameters if we are to understand the structures of molecules and their properties, particularly their photochemical reactions. As in the preceding chapters, we begin by describing a system that can be solved exactly: the hydrogen atom. Then we build on our knowledge of that atom’s structure and spectra to discuss the properties and structures of many-electron atoms.
Many-electron atoms 7.12 Penetration and shielding 7.13 Periodicity 7.14 Slater atomic orbitals
The spectrum of atomic hydrogen
7.15 Self-consistent fields 7.16 Term symbols and transitions of many-electron atoms 7.17 Hund’s rules and the relative energies of terms 7.18 Alternative coupling schemes
So long as we ignore electron spin, the state of an electron in a hydrogen atom is specified by three quantum numbers, n, l, and ml (Section 3.11) and its energy is given by En ¼
Atoms in external fields 7.19 The normal Zeeman effect
me4 32p2 e20 h2
!
1 n2
n ¼ 1, 2, . . .
ð7:1Þ
This expression is normally written
7.20 The anomalous Zeeman effect 7.21 The Stark effect
En ¼
hcRH n2
RH ¼
me4 8e20 h3 c
ð7:2Þ
where RH is the Rydberg constant for hydrogen. The origin of this expression was explained in Chapter 3 and there is no need to repeat the arguments here, but for convenience the array of energy levels is shown in Fig. 7.1.
208
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
7.1 The energies of the transitions
0 –1/16 –1/9
s
p
d
f 3
–1/4
∞ 4
Energy/hcRH
2
The spectrum of atomic hydrogen arises from transitions between its permitted states, and the difference in energy, DE, between the states is discarded as a photon of energy hn and wavenumber ~n, where ~n ¼ n=c. For the transition n2 ! n1, the wavenumber of the emitted radiation is 1 1 ~n ¼ RH ð7:3Þ n21 n22 For a given value of n1, the set of transitions from n2 ¼ n1 þ 1, n1 þ 2, . . . constitutes a series of lines, and these series bear the names of their discoverers or principal investigators: n1 ¼ 1, n2 ¼ 2, 3, . . . Lyman series, ultraviolet n1 ¼ 2, n2 ¼ 3, 4, . . . Balmer series, visible n1 ¼ 3, n2 ¼ 4, 5, . . . Paschen series, infrared n1 ¼ 4, n2 ¼ 5, 6, . . . Brackett series, far infrared
–1
1
Fig. 7.1 The energy levels of
the hydrogen atom. Hydrogenic atoms in general have the same spectrum, but with the energy scale magnified by a factor of Z2.
n1 ¼ 5, n2 ¼ 6, 7, . . . Pfund series, far infrared n1 ¼ 6, n2 ¼ 7, 8, . . . Humphreys series, far infrared Because each series corresponds to a specific value of n1 but all possible integer values of n2 (provided n2 > n1), the limit of each series is the wavenumber obtained by setting n2 ¼ 1 in eqn 7.3, and is given by ~n1 ¼
RH n21
ð7:4Þ
The energy when n ¼ 1 is zero (eqn 7.1), and corresponds to the complete removal of the electron from the atom; that is, n ¼ 1 corresponds to the ionized state of the atom. The ionization energy I of the atom, the minimum energy required to ionize it from its n ¼ 1 ground state, is the energy difference E1 E1. Hence, I ¼ hcRH
ð7:5Þ 18
The numerical value of the ionization energy is 2.180 aJ (where 1 aJ ¼ 10 which corresponds to 1312 kJ mol1 and 13.60 eV.
J),
Example 7.1 Determining the ionization energy of a hydrogenic atom
What is the ionization energy of Heþ? Method. The energy for one-electron ions is given by eqn 3.44. To a good
approximation, we can use me in place of m and, subsequently, use the value of R given on the inside front cover. The ground state of the ion is n ¼ 1; for He, Z ¼ 2. Answer. The ionization energy is the energy difference E1 E1 and is
I ¼ hcZ2 R ¼ 8:719 aJ
7.2 SELECTION RULES
j
209
Comment. For greater accuracy, we should use RHe, and take into account the reduced mass of the electron and the helium nucleus. The ionization energy of Heþ is also the second ionization energy (the energy required to remove a second electron from the ground-state species) of neutral He. Self-test 7.1. What is the ionization energy of Li2þ?
7.2 Selection rules Not all transitions between states are allowed. The selection rules for electricdipole transitions, the rules that specify the specific transitions that may occur, are based on an examination of the transition dipole moment (Section 6.17) between the two states of interest. They are established by identifying the conditions under which the transition dipole moment is non-zero, corresponding to an allowed transition, or zero, for a forbidden transition. The transition dipole moment for a transition between states jii and jfi is defined as if ¼ hijjfi
ð7:6Þ
where ¼ er is the electric dipole operator. The transition dipole moment can be regarded as a measure of the size of the electromagnetic jolt that the electron delivers to the electromagnetic field when it makes a transition between states. Large shifts of charge through large distances can deliver strong impulses provided they have a dipolar character (as in the transition between s- and p-orbitals but not between s-orbitals where the shift of charge is spherically symmetrical), and such transitions give rise to intense lines. Group theory (Section 5.16) tells us that a transition dipole moment must be zero unless the integrand in eqn 7.6 is totally symmetric under the symmetry operations of the system, which for atoms is the full rotation group, R3. The easiest operation to consider is inversion, under which r ! r. Under inversion, an atomic orbital with quantum number l has parity (1)l, as can be appreciated by noting that orbitals with even l (s- and d-orbitals for example) do not change sign whereas those with odd l (p- and f-orbitals for example) do change sign. This behaviour is also apparent from the mathematical form of the spherical harmonics (see Table 3.1). The parity of the integrand is therefore (1)li(1)(1)lf, which is even if the two orbitals have opposite parity (one odd, the other even). This argument is the basis of the Laporte selection rule: The only allowed electric-dipole transitions are those involving a change in parity. Next, consider the rotational characteristics of the components of the integrand. The atomic orbitals are angular momentum wavefunctions and span the irreducible representations G(li) and G(lf) of the full rotation group. The electric dipole moment operator behaves like a translation and, recalling the relation among l ¼ 1 spherical harmonics and Cartesian coordinates,
210
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE 1 l
l
spans the irreducible representation G(1) of the group. The product of G(li) and G(1) therefore spans Gðli Þ Gð1Þ ¼ Gðli þ1Þ þ Gðli Þ þ Gðli 1Þ
l +1
1
l
l
l –1
as explained in Section 5.20. For the product of all three factors in the integrand to span the totally symmetric irreducible representation (G(0)), we require G(lf) to be equal to G(li þ 1), G(li), or G(li 1). In other words, lf ¼ li 1, li, or li þ 1. However, we have already ruled out transitions that do not change parity, so the only allowed transitions are those to the states with lf ¼ li 1. That is: Dl ¼ 1
Fig. 7.2 The basis of the Dl ¼ 1 selection rule is the conservation of angular momentum and the fact that a photon has a helicity (the projection of its spin on its direction of propagation) of 1. Note that the absorption of a photon (as depicted in both instances here) can result in either an increase or a decrease of l.
= –1 ml = –1 z
= +1 ml = +1
ð7:7Þ
The origin of this selection rule can be put on a more physical basis by noting that the intrinsic spin angular momentum of a photon is 1.1 Therefore, when it is absorbed or emitted, to conserve total angular momentum, the orbital angular momentum of the electron in the atom must change by 1. An increase in orbital angular momentum (Dl ¼ þ1) can accompany either an absorption or an emission of a photon, depending on the orientation of the angular momentum of the photon relative to the angular momentum of the electron in the atom (Fig. 7.2). It is quite easy to extend these pictures to obtain the selection rules for ml, the magnetic orbital quantum number. Now we need to know that a photon has an intrinsic helicity, s, the spin angular momentum relative to its line of flight, of s ¼ 1 (Fig. 7.3). We shall suppose that ml labels the component of orbital angular momentum on the axis defined by the line of flight of the photon. Then, absorption of a left-circularly polarized photon (with helicity s ¼ þ1) results in Dml ¼ þ1 to preserve overall angular momentum, and its emission results in Dml ¼ 1. The opposite holds for a right-circularly polarized photon. The maximum change in ml is therefore 1. It follows that for an atom that has its electron with a definite value of ml for the component of angular momentum relative to an arbitrary axis, not necessarily the line of flight of the photon, the maximum change in ml is still 1 but an allowed intermediate value may also occur if the photon is travelling in an intermediate direction. Therefore, the general selection rule is Dml ¼ 0, 1
Fig. 7.3 The change in ml that accompanies the absorption of a photon; the z-axis is taken to be the line of flight of the photon.
ð7:8Þ
The selection rule on ml can also be deduced algebraically. Suppose the radiation is plane-polarized with the electric field in the z-direction, then only the z-component of the dipole moment is relevant, and we can write mz ¼ er cosy. The f integral in the transition moment is then proportional to Z 2p Z 2p eimlf f ðer cos yÞeimli f df / eiðmli mlf Þf df 0
0
.......................................................................................................
1. A photon, having integral spin, is a boson (Section 7.11).
7.2 SELECTION RULES
j
211
The integral over f is zero unless mli ¼ mlf. Therefore, for z-polarized radiation, Dml ¼ 0. The selection rules Dml ¼ 1 arise similarly for radiation polarized in the xy-plane. Example 7.2 The calculation of transition moments
Calculate the electric dipole transition moment for the transition 2pz ! 2s in a hydrogenic atom. Method. We use the wavefunctions set out in Tables 3.1 and 3.2 to evaluate
the integral h2pz j mz j 2si with mz ¼ er cos y. Answer. The wavefunctions for the orbitals are
c2pz ¼ c2s ¼
Z5 32pa50 Z3 32pa50
1=2 1=2
r cos yeZr=2a0 ð2a0 ZrÞeZr=2a0
The integral we require is therefore 4 Z 1 Z p Z 2p Z 4 Zr=a0 2 2pz mz 2s ¼ e ð2a ZrÞr e dr cos ysinydy df 0 32pa50 0 0 0 3ea0 ¼ Z For the hydrogen atom itself, h2pz jmz j2si¼ 3ea0. Comment. The sign of the transition dipole moment has no physical sig-
nificance because the relative signs of the wavefunctions used to calculate it are arbitrary. The physical observable, the transition intensity, depends on the square modulus of the transition dipole moment. Self-test 7.2. Repeat the calculation for the transition 2pz ! 1s.
Electric dipole transitions are not the only types of transition that may occur. Light is an electromagnetic phenomenon, and the perturbation arising from the effect of the magnetic component of the field can induce magnetic dipole transitions. Such transitions have intensities that are proportional to the squares of matrix elements like hijlzjfi and are typically about 105 times weaker than allowed electric dipole transitions. However, because they obey different selection rules, they may give rise to spectral lines where the electric dipole transition is forbidden. Another type of transition is an electric quadrupole transition in which the spatial variation of the electric field interacts with the electric quadrupole moment operator. Such transitions have intensities that are proportional to the squares of matrix elements like hijxyjfi. These transitions are about 108 times weaker than electric dipole transitions. Their selection rule is Dl ¼ 0, 2. The large change in angular momentum that accompanies the transition arises from the fact that the quadrupole transition imparts an orbital angular momentum to the photon
212
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
(that is, generates it with a non-spherically symmetric wavefunction) in addition to its intrinsic spin. The weakness of magnetic dipole and electric quadrupole transitions stems from the fact that both depend on the variation of the electromagnetic wave over the extent of the atom. As atomic diameters are much smaller than typical wavelengths of radiation, this variation is typically very small and the intensity is correspondingly weak. In some systems, a transition can result in the generation of two photons by an electric dipole mechanism more efficiently than a single photon is generated by a magnetic dipole transition. An example of this multiple-quantum dipole transition is provided by the excited 1s12s1 singlet state of helium: the two-photon process governs the lifetime of the state because the magnetic dipole transition probability is so low.
7.3 Orbital and spin magnetic moments So far, we have ignored the spin of the electron. Now we consider its effect on the structure and spectra of hydrogenic atoms. Its effect is not very pronounced on the energy levels of hydrogen itself, but it can be of great importance for atoms of high atomic number. We note that an electron has spin quantum number s ¼ 12 and that the spin magnetic quantum number is one of the two values ms ¼ 12. An electron is a charged particle and there is a magnetic moment associated with its angular momentum. Because the electron in an atom may have two types of angular momentum, spin and orbital angular momentum, there are two sources of magnetic moment. These two magnetic moments can interact and give rise to shifts in the energies of the states of the atom which affect the appearance of the spectrum of the atom. The resulting shifts and splitting of lines is called the fine structure of the spectrum. First, consider the magnetic moment arising from the orbital angular momentum of the electron. The quantum mechanical derivation of its orbital magnetic moment is described in Section 13.6; here we shall use the following classical argument. If a particle of charge e circulates in an orbit of radius r in the xy-plane at a speed v, the current generated is ev I¼ 2pr This current gives rise to a magnetic dipole moment with z-component mz ¼ IA, where A is the area enclosed by the orbit, A ¼ pr2. It follows that mz ¼ IA ¼
evpr2 ¼ 12evr 2pr
The z-component of the orbital angular momentum of the electron is lz ¼ mevr (recall l ¼ r p and p ¼ mv), so e mz ¼ lz 2me The same argument applies to orbital motion in other planes, and we can therefore write m ¼ ge l
ð7:9Þ
7.3 ORBITAL AND SPIN MAGNETIC MOMENTS
j
213
where ge ¼
e 2me
ð7:10Þ
The constant ge is called the magnetogyric ratio of the electron. The properties of the orbital magnetic moment m follow from those of the angular momentum itself. In particular, its z-component is quantized and restricted to the values mz ¼ ge ml h ml ¼ l, l 1, . . ., l
ð7:11Þ
The positive quantity mB ¼ ge h¼
e h 2me
ð7:12Þ
is called the Bohr magneton, and is often regarded as the elementary unit of magnetic moment. Its value is 9.274 10 24 J T1. In terms of the Bohr magneton, the z-component of orbital magnetic moment is mz ¼ mB ml Now we consider the magnetic moment that arises from the spin of the electron. By analogy with the orbital magnetic moment, we might expect the spin magnetic moment to be related to the spin angular momentum by m ¼ ges, but this turns out not to be the case. This should not be too surprising however, because spin has no classical analogue, yet here we are trying to argue by analogy with orbital angular momentum, which does have a classical analogue. The relation between the spin and its magnetic moment can be derived from the relativistic Dirac equation, which gives m ¼ 2ges: the magnetic moment due to spin is twice the value expected on the basis of a classical analogy. The experimental value of the magnetic moment can be determined by observing the effect of a magnetic field on the motion of an electron beam, and it is found that m ¼ ge ge s
where ge ¼ 2:002 319 304
ð7:13Þ
The factor ge is called the g-factor of the electron. The small discrepancy between the experimental value and the Dirac value of exactly 2 is accounted for by the more sophisticated theory of quantum electrodynamics, in which charged particles are allowed to interact with the quantized electromagnetic field.2 As for the orbital magnetic moment, the spin magnetic moment has quantized components on the z-axis, and we write mz ¼ ge mB ms
ms ¼ 12
ð7:14Þ
.......................................................................................................
2. The following classical picture might be helpful. Quantum electrodynamics expresses the electromagnetic field as a collection of harmonic oscillators. We have seen that a harmonic oscillator has a zero-point energy, and so the electromagnetic vacuum has fluctuating electric and magnetic fields even if no photons are present. These vacuum fluctuations interact with the electron, and instead of moving smoothly the electron jitterbugs (technically, this motion is called Zitterbewegung). It also wobbles as it spins (in so far as spin has any such significance), for the same reason, and the wobble increases its magnetic moment above the value that would be expected for a smoothly spinning object.
214
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
7.4 Spin–orbit coupling We now turn to the energy of interaction between the two magnetic moments, spin and orbital, of an electron. In fact, we shall use this opportunity to emphasize the danger of arguing by classical analogy, particularly when spin is involved. The classical calculation of the energy of interaction runs as follows. A particle of mass me and charge e moving at a velocity v in an electric field E experiences a magnetic field B¼
E v c2
If the field is due to an isotropic electric potential f, we can write E¼
r df r dr
It follows that B¼
1 df r v rc2 dr
The orbital angular momentum of the particle is l ¼ r p ¼ mer v, and so B¼
1 df l me rc2 dr
ð7:15Þ
The energy of interaction between a magnetic field B and a magnetic dipole m is m B, so we might anticipate (using eqns 7.10, 7.13, and 7.15 and taking ge ¼ 2) that the spin–orbit coupling hamiltonian should be Hso ¼ m B ¼
1 df e df ml ¼ 2 2 sl me rc2 dr me rc dr
ð7:16Þ
It turns out that this is exactly twice the result obtained by solving the Dirac equation. The error in the above formulation is the implicit assumption that one can step from the stationary nucleus to the moving electron without treating the change of viewpoint relativistically.3 The correct calculation gives Hso ¼ xðrÞl s
ð7:17Þ
where x (xi) is given by xðrÞ ¼
e df 2m2e rc2 dr
ð7:18Þ
.......................................................................................................
3. The phenomenon that gives rise to the factor 12 is called Thomas precession. The electron moves in its orbital with speeds that approach the speed of light. To an observer on the nucleus, the coordinate system seems to rotate in the plane of motion, and the electron moves in such a way that its coordinate system appears to rotate by 180 when it has completed one circuit of the nucleus. It is spinning (in a classical sense) within its own frame with only one-half the rate if the frame were stationary, and this virtual slowing of its apparent motion reduces its magnetic moment by a factor of 12.
7.4 SPIN–ORBIT COUPLING
j
215
The radial average for the state jnlmli of the function x(r) h2is written hcz, where z (zeta) is called the spin–orbit coupling constant; specifically h2 hcznl ¼ hnlml jxðrÞjnlml i
ð7:19Þ
The same value is obtained regardless of the value of ml because the electric potential is isotropic. Defined in this way, z is a wavenumber and hcz is an energy. For an electron in a hydrogenic atom, the potential arising from a nucleus of charge Ze is Coulombic, and f¼
Ze 4pe0 r
Consequently xðrÞ ¼
Ze2 8pe0 m2e r3 c2
ð7:20Þ
The expectation value of r 3 for hydrogenic orbitals, using the general properties of associated Laguerre functions (Section 3.11), is hnlml jr3 jnlml i ¼
Z3 n3 a30 lðl þ 12Þðl þ 1Þ
ð7:21Þ
where a0 is the Bohr radius (eqn 3.43). Therefore, the spin–orbit coupling constant for a hydrogenic atom is hcznl ¼
Z4 e 2 h2 8pe0 m2e c2 n3 a30 lðl þ 12Þðl þ 1Þ
ð7:22Þ
It proves useful to express this ungainly formula in terms of the fine-structure constant, a, which is defined as a¼
e2 4pe0 hc
ð7:23Þ
This dimensionless collection of fundamental constants has a value close to 1/137 (more precisely, a ¼ 7.297 35 10 3) and is of extraordinarily broad significance because it is a fundamental constant for the strength of the coupling of a charge to the electromagnetic field. In the present context, we can use it to write znl ¼
a2 RZ4 n3 lðl þ 12Þðl þ 1Þ
ð7:24Þ
where R is the Rydberg constant obtained by replacing m in eqn 7.2 by me (see inside front cover). For hydrogen itself, Z ¼ 1, and for a 2p-electron z ¼ a2R/24, which is about 2.22 10 6 R. Energy level separations and the wavenumbers of transitions (see eqn 7.3) are of the order of R itself, so the fine structure of the spectrum of atomic hydrogen is a factor of about 2 10 6 times smaller, or of the order of 0.2 cm1, as observed. In passing, note that as z / Z4, spin– orbit coupling effects are very much larger in heavy atoms than in light atoms. What may be seen as a niggling problem in hydrogen can be of dominating
216
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
importance in heavy elements, and the work we are doing here will prepare us for them. High energy
l High j s
(a)
l Low j
(b)
Low energy
s
Fig. 7.4 (a) High and (b) low energy relative orientations of spin and orbital angular momenta of an electron as a result of the interaction of the corresponding angular momenta. The black arrows denote angular momenta and the blue arrows denote magnetic moments.
7.5 The fine-structure of spectra We can now explore how the spin–orbit coupling affects the appearance of spectra. Consider Fig. 7.4. When the spin and orbital angular momenta are parallel, the total angular momentum quantum number, j, takes its highest value (j ¼ 32 for l ¼ 1 and s ¼ 12, and l þ 12 in general). The corresponding magnetic moments are also parallel, which is a high-energy arrangement (eqn 7.17). When the two angular momenta are antiparallel, j has its minimum value (j ¼ 12 when l ¼ 1 and s ¼ 12, and j ¼ l 12 in general). The corresponding magnetic moments are now antiparallel, which is a low-energy arrangement. We conclude that the energy of the level with j ¼ l þ 12 should lie above the level with j ¼ l 12, and that the separation should be of the order of the spin–orbit coupling constant as that is a measure of the strength of the magnetic interaction between momenta. Note that the high energy of a state with high j does not stem directly from the fact that the total angular momentum is high, but rather stems from the fact that a high j indicates that two magnetic moments are parallel and hence interacting adversely. Without that interaction, high j and low j would have the same energy. Because the spin–orbit interaction is so weak in comparison with the energy-level separations of the atom, we can use first-order perturbation theory to assess its effect. The first-order correction to the energy of the state j ls;jmji is Eso ¼ hls; jmj jHso jls; jmj i ¼ hls; jmj jxðrÞl sjls; jmj i
ð7:25Þ
(In the language of Section 4.9, note that we are using the coupled representation of the state, which is the natural one to use for the problem.) The matrix elements of a scalar product can be evaluated very simply by noting that j2 ¼ jl þ sj2 ¼ l2 þ s2 þ 2l s
ð7:26Þ
Therefore, l sjls; jmj i ¼ 12ð j2 l2 s2 Þjls; jmj i ¼ 12 h2 f jð j þ 1Þ lðl þ 1Þ sðs þ 1Þgjls; jmj i
ð7:27Þ
Consequently, the interaction energy is Eso ¼ 12 h2 f jð j þ 1Þ lðl þ 1Þ sðs þ 1Þghls; jmj jxðrÞjls; jmj i ¼ 12hcznl f jð j þ 1Þ lðl þ 1Þ sðs þ 1Þg ( ) jð j þ 1Þ lðl þ 1Þ sðs þ 1Þ 4 2 ¼ Z a hcR 2n3 lðl þ 12Þðl þ 1Þ
ð7:28Þ
Note that the energy is independent of mj, the orientation of the total angular momentum in space, as is physically plausible, so each level is (2j þ 1)-fold degenerate. The matrix element hls;jmj j x j ls;jmji is independent of s, j, and mj
7.6 TERM SYMBOLS AND SPECTRAL DETAILS
j= l = 1, s =
3 2
j
217
because x depends only on the radius r; as a result, the matrix element may be identified with hcznl/h2. For an s-electron, the spin–orbit interaction is zero because the electron has no orbital angular momentum. Specifically, because j ¼ s when l ¼ 0, h2 fsðs þ 1Þ 0 sðs þ 1Þg ¼ 0 h0s; sms jl sj0s; sms i ¼ 12
1 2
j=
1 2
Fig. 7.5 The splitting of the states of a p-electron by spin–orbit coupling. Note that the centre of gravity of the levels is unshifted.
For a p-electron, the separation between levels with j ¼ 32 and j ¼ 12 is Z4a2hcR/2n3, and so it rapidly becomes negligible as n increases. For a hydrogen 2p-electron the splitting is a2R/16 0.365 cm1. It should be noted that the centroid of the split levels, with each one weighted by its degeneracy, is at the same energy as the unsplit state (Fig. 7.5), as illustrated below.
Illustration 7.1 Finding the centroid of split levels
For a p-electron j ¼ 32 or 12. We need to focus only on the term j(j þ 1) l(l þ 1) s(s þ 1) from eqn 7.28 as all other terms for Eso are fixed for specified values of n and l. For j ¼ 32, there are 2j þ 1 ¼ 4 degenerate states and they are raised in energy (relative to l ¼ 1, s ¼ 12) by an amount proportional to j(j þ 1) l(l þ 1) s(s þ 1) ¼ 1. Similarly, for j ¼ 12, there are 2j þ 1 ¼ 2 degenerate states and their increase in energy is proportional to j(j þ 1) l(l þ 1) s(s þ 1) ¼ 2; that is, they are lowered in energy. The centroid of the split levels remains unchanged: 4 1 þ 2 (2) ¼ 0.
7.6 Term symbols and spectral details To simplify the discussion of the spectrum that arises from these energy levels we need to introduce some more notation. Spectral lines arise from transitions between terms, which is another name for energy levels. The wavenumber, ~n, of a transition is the difference between the energies of two terms expressed as wavenumbers: ~n ¼ T 0 T
ð7:29Þ 0
0
A transition is denoted T ! T for emission and T T for absorption, with the term T 0 higher in energy than the term T. The configuration of an atom is the specification of the orbitals that the electrons occupy. There is only one electron in hydrogen, so we speak of the configuration 1s1 if the electron occupies a 1s-orbital, 2s1 if it occupies a 2sorbital, and so on.4 A single configuration (such as 2p1) may give rise to several terms. For hydrogen, each configuration with l > 0 gives rise to a doublet term in the sense that each term splits into two levels with different values of j, namely j ¼ l þ 12 and j ¼ l 12. For example, the configuration 2p1 gives rise to a doublet term with the levels j ¼ 32 and j ¼ 12, the configuration 3d1 gives rise to a doublet term with the levels j ¼ 52 and j ¼ 32, and so on. Each level .......................................................................................................
4. For hydrogen, 1s1 is the ground-state configuration; all others, such as 2s1, are excited-state configurations.
States
Levels
Terms
labelled by the quantum number j consists of 2j þ 1 individual states distinguished by the quantum number mj. The hierarchy of concepts is summarized in Fig. 7.6. The level of each term arising from a particular configuration is summarized by a term symbol:
Split by external fields
Split by magnetic interactions
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE Split by electrostatic interactions
j
Configuration
218
Fig. 7.6 The hierarchy of names and the origin of the splittings that occur in atoms.
1 2
3s
1 2
S1/2
3p
3d P3/2
12
3d
D5/2 12
D3/2
12
3p P1/2 H 1 2
2p 2s
1 2
S1/2
1 2
2p
P3/2
P1/2
1s S1/2 1 2
Fig. 7.7 The energy levels of a hydrogen atom showing the fine structure and the transitions that give rise to certain features in the spectrum. Note that in this approximation some degeneracies remain (for states of the same j).
multiplicity!2Sþ1
fLg
J
Level
orbital angular momentum
where {L} is a letter (S, P, D, F, etc.) corresponding to the value of the total orbital angular momentum quantum number L (0, 1, 2, 3, etc.). For a hydrogen atom, L ¼ l, so a configuration ns1 gives rise to an S term, a configuration np1 gives rise to a P term, and so on. The multiplicity of a term is the value of 2S þ 1, where S is the total spin angular momentum quantum number; provided that L S, the multiplicity is the number of levels of the term. For hydrogen, S ¼ s ¼ 12, so 2S þ 1 ¼ 2, and all terms are doublets and are denoted 2 S, 2P, etc. As we saw earlier, all terms other than 2S have two levels distinguished by the value of J, and for hydrogen J ¼ j. A 2S term has only a single level, with J ¼ j ¼ s ¼ 12. The precise level of a term is specified by the right subscript of the term symbol, as in 2S1/2 and 2P3/2. Each of these levels consists of 2J þ 1 states, but these are rarely specified in a term symbol as they are degenerate in the absence of external electric and magnetic fields.
7.7 The detailed spectrum of hydrogen The transitions responsible for the spectrum of hydrogen can be expressed using term symbols (Fig. 7.7). Consider, for instance, the transitions responsible for the Ha line in the Balmer series (the line responsible for the red glow of excited hydrogen atoms). The upper terms have n ¼ 3 and the lower have n ¼ 2. The configuration 3s1 gives rise to a 2S1/2 term with a single level. The configuration 3p1 gives rise to 2P3/2 and 2P1/2, with a very small spin– orbit splitting between the two levels. The 3d1 configuration gives rise to the levels 2D5/2 and 2D3/2. In each case, the level with the lower value of J lies lower in energy. The configuration 2s1 similarly gives rise to a term 2S1/2 and the configuration 2p1 gives rise to 2P3/2 and 2P1/2 with a splitting of about 0.36 cm1, as explained before. One possibly confusing point is that, according to the Dirac theory of the hydrogen atom, the energy of the ns1 2S1/2 term is the same as that of the np1 2 P1/2 term (see Fig. 7.7). One way to view this degeneracy is that the Schro¨dinger equation ignores relativistic effects. When these effects are taken into account (as they are by the Dirac equation), they give rise to a contribution to the energy which is of the same order of magnitude as the spin–orbit interaction (which is also a relativistic phenomenon), with the result that levels of the same value of j but different values of l are degenerate. Nevertheless, although the Dirac equation predicts an exact degeneracy, there is experimentally a small splitting between 2S1/2 and 2P1/2, which is known as the Lamb shift. As in the case of other discrepancies between experiment and the Dirac equation, we have to look for an explanation in the role of the electromagnetic vacuum in which the atom is immersed, and quantum
7.8 THE HELIUM ATOM
j
219
electrodynamics accounts fully for the Lamb shift. The pictorial explanation appeals to the role of the zero-point fluctuations of the oscillations of the electromagnetic field, and their influence on the motion of the electron. This jitterbugging motion of the electron tends to smear its location over a region of space. The effect of this smearing on the energy is most pronounced for s-electrons, as they spend a high proportion of their time close to the nucleus. The smearing tends to reduce the probability that the electron will be found at the nucleus itself, and so the energy of the orbital is raised slightly. There is less effect on the energy of a p-electron because it spends less time close to the nucleus and its interaction with the nucleus is less sensitive to the smearing. The allowed transitions between terms arising from the configurations with n ¼ 3, 2, and 1 are shown in Fig. 7.7 (the selection rules on which this illustration is based are discussed later). Because the only appreciable spin–orbit splitting occurs in the 2p1configuration, the transitions contributing to the Ha line fall into two groups separated by 0.36 cm1. The doublet structure in the spectrum is therefore a compound doublet arising from two almost coincident groups of transitions.
The structure of helium We now move towards a discussion of many-electron atoms by setting up an approximate description of the simplest example: the helium atom. We shall then use the features that this atom introduces to discuss more complex atoms.
7.8 The helium atom The hamiltonian for the helium atom (Z ¼ 2) is H¼ r1
r12 r2
Fig. 7.8 The distances involved in the potential energy of a two-electron atom.
2 h 2e2 2e2 e2 ðr21 þ r22 Þ þ 2me 4pe0 r1 4pe0 r2 4pe0 r12
ð7:30Þ
with the distances defined in Fig. 7.8. The first two terms are the kinetic energy operators for the two electrons, the following two are the potential energies of the two electrons in the field of the nucleus of charge 2e, and the final term is the potential energy arising from the repulsion of the two electrons when they are separated by a distance r12. In a very precise calculation we should use the reduced mass of the electron, but the calculation will be so crude that this refinement is unnecessary. The Schro¨dinger equation has the form Hcðr 1 , r 2 Þ ¼ Ecðr 1 , r 2 Þ
ð7:31Þ
and the wavefunction depends on the coordinates of both electrons. It appears to be impossible to find analytical solutions of such a complicated partial differential equation in six variables (this is due to the presence of the electronic repulsion term in eqn 7.30), and almost all work has been directed
220
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
towards finding increasingly refined numerical solutions. The simplest version of these approximate solutions is based on a perturbation approach, and this is the line we shall initially take here. The obvious candidate to use as the perturbation is the electron–electron interaction, but as it is not particularly small compared with the other terms in the hamiltonian we should not expect very good agreement with experiment, and will need to make further refinements. The unperturbed system is described by a hamiltonian that is the sum of two hydrogenic hamiltonians: Hð0Þ ¼ H1 þ H2 ,
Hi ¼
2 2 h 2e2 ri 2me 4pe0 ri
ð7:32Þ
Whenever a hamiltonian is expressed as the sum of two independent terms, the eigenfunction is the product of two factors. Proof 7.1 Eigenfunction as a product of independent factors
We seek the eigenfunction c such that H(0)c ¼ Ec, with H(0)a sum of independent terms H1 þ H2 þ þ Hn. Writing c as a product of independent terms c1c2 . . . cn, where Hici ¼ Eici, we find ðH1 þ H2 þ þ Hn Þc ¼ ðH1 þ H2 þ þ Hn Þc1 c2 . . . cn ¼ ðH1 c1 Þc2 . . . cn þ c1 ðH2 c2 Þc3 . . . cn þ þ c1 c2 . . . cn1 ðHn cn Þ ¼ E1 c1 c2 . . . cn þ E2 c1 c2 . . . cn þ þ En c1 c2 . . . cn ¼ ðE1 þ E2 þ þ En Þc Therefore, the product c ¼ c1c2 . . . cn is an eigenfunction with eigenvalue E ¼ E1 þ E2 þ þ En.
It follows that for helium the wavefunction of the two electrons (with their repulsion disregarded) is the product of two hydrogenic wavefunctions:5 cðr 1 , r 2 Þ ¼ cn1 l1 ml1 ðr 1 Þcn2 l2 ml2 ðr 2 Þ and that, from eqn 3.44, the energies are 1 1 E ¼ 4hcR 2 þ 2 n1 n2
ð7:33Þ
ð7:34Þ
where we use R (inside front cover) because we are replacing the true reduced mass with the electron mass. .......................................................................................................
5. This simple product of two hydrogenic wavefunctions is most appropriate when both electrons occupy the same orbital, as in the ground state of He. When electrons occupy different orbitals, see Section 7.9.
j
7.8 THE HELIUM ATOM
221
Now consider the influence of the electron–electron repulsion term. The first-order correction to the energy is e2 n1 l1 ml1 ; n2 l2 ml2 i ¼ J ð7:35Þ Eð1Þ ¼ hn1 l1 ml1 ; n2 l2 ml2 4pe r 0 12
Electron density in orbital 1
J
Electron density in orbital 2 Fig. 7.9 The physical interpretation
of the Coulomb integral, J.
The term J is called the Coulomb integral: Z e2 1 jcn2 l2 ml2 ðr 2 Þj2 dt1 dt2 jcn1 l1 ml1 ðr 1 Þj2 J¼ r12 4pe0
ð7:36Þ
This integral (which is positive) has a very simple interpretation (Fig. 7.9). The term jcn1 l1 ml1 ðr 1 Þj2 dt1 is the probability of finding the electron in the volume element dt1, and when multiplied by e it is the charge associated with that region. Likewise, ejcn2 l2 ml2 ðr 2 Þj2 dt2 is the charge associated with the volume element dt2. The integrand is therefore the Coulombic potential energy of interaction between the charges in these two volume elements and J is the total contribution to the potential energy arising from electrons in the two orbitals. Example 7.3 Evaluation of a Coulomb integral
Evaluate the Coulomb integral for the configuration 1s2 of a hydrogenic atom given the following expansion6 l 1 1 X 4p r2 ¼ Ylm ðy1 , f1 ÞYlml ðy2 , f2 Þ l r12 r1 l;m 2l þ 1 r1 l
when rl > r2, and with r1 and r2 interchanged when r1 < r2. Method. The integral should be evaluated using c ¼ ðZ3 =pa30 Þ
1=2 Zrla0
e for each electron. Because the wavefunctions are independent of angle, the integration over the angles is straightforward: the integration over Y gives zero except when l ¼ 0 and ml ¼ 0. Hence, the sum given above reduces to a single term inside the integral, namely 1/r12 ¼ 1/r1 when r1 > r2 and 1/r12 ¼ 1/r2 when r2 > r1. The radial integrations should be divided into two parts, one with r1 > r2 and the other with r2 > r1. Answer. The integration is as follows:
3 2 Z 2p Z 2p Z p Z p e2 Z df df sin y dy sin y2 dy2 1 1 1 2 4pe0 pa3 0 0 0 0 Z 1 Z 1 02Zðr1 þr2 Þ=a0 e r21 r22 dr1 dr2 r12 0 0 2 3 2 Z 1 Z r2 2 2Zr1 =a0 e Z r1 e 2 ¼ ð4pÞ dr1 4pe0 r2 pa30 0 0
Z 1 2 2Zr1 =a0 r1 e dr1 r22 e2Zr2 =a0 dr2 þ r1 r2 2 3 2 e Z 5 a0 5 5 e2 Z 2 ¼ ð4pÞ ¼ 27 Z 8 4pe0 a0 4pe0 pa30
J¼
.......................................................................................................
6. See, for example, eqn (3.70) of J.D. Jackson, Classical electrodynamics, Wiley (1975).
222
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
For helium, Z ¼ 2, and so J¼
2 5 e 5:45 aJ 4 4pe0 a0
Comment. Take care with the expansion when orbitals other than s-orbitals
are involved, because additional terms then survive. Self-test 7.3. Evaluate J for the configuration 1s12s1.
It is shown in the example that J 5.45 aJ, which corresponds to 34 eV or 2.50hcR. The total energy of the ground state of the atom in this approximation is therefore E ¼ ð4 4 þ 2:50ÞhcR ¼ 5:50hcR This value corresponds to 12.0 aJ, or 7220 kJ mol1. The experimental value, which is equal to the sum of the first and second ionization energies of the atom, is 7619 kJ mol1 (12.65 aJ, 5.804hcR). The agreement is not brilliant, but the calculation is obviously on the right track. One of the reasons for the disagreement is that the perturbation is not small, and so first-order perturbation theory cannot be expected to lead to a reliable result.
7.9 Excited states of helium A new feature comes into play when we consider the excited states of the atom. When the two electrons occupy different orbitals (as in the configuration 1s12s1), the wavefunctions are either cn1l1ml1(r1)cn2l2ml2(r1) or cn2l2ml2(r1) cnll1ml1(r2), which we shall denote a(1)b(2) and b(1)a(2), respectively. Both wavefunctions have the same energy and their unperturbed energies are Ea þ Eb. To calculate the perturbed energy, we use the form of perturbation theory appropriate to degenerate states (Section 6.8), and therefore set up the secular determinant. To do so, we need the following matrix elements, in which we identify state 1 with a(1)b(2) and state 2 with b(1)a(2): e2 að1Þbð2Þi ¼ Ea þ Eb þ J H11 ¼ hað1Þbð2ÞH1 þ H2 þ 4pe0 r12 H22 ¼ Ea þ Eb þ J H12 ¼ hað1Þbð2ÞH1 þ H2 þ
e2 að2Þbð1Þi 4pe0 r12
¼ ðEa þ Eb Þhað1Þbð2Þjað2Þbð1Þi þ hað1Þbð2Þ
e2 að2Þbð1Þi ¼ H21 4pe0 r12
The first of the integrals in H12 is zero because the orbitals a and b are orthogonal: hað1Þbð2Þjað2Þbð1Þi ¼ hað1Þjbð1Þihbð2Þjað2Þi ¼ 0
7.9 EXCITED STATES OF HELIUM
The remaining integral is called the exchange integral, K: 1 e2 hað1Þbð2Þ að2Þbð1Þi K¼ r12 4pe0
j
223
ð7:37Þ
Like J, this integral is positive. The secular determinant is therefore H11 ES11 H12 ES12 H11 E H12 ¼ H21 ES21 H22 ES22 H21 H22 E Ea þ Eb þ J E K ¼ K Ea þ Eb þ J E ¼0
ð7:38Þ
(Note that S11 ¼ S22 ¼ 1 and S12 ¼ S21 ¼ 0 due to orthonormality of states 1 and 2.) The solutions are E ¼ Ea þ Eb þ J K
ð7:39Þ
and the corresponding wavefunctions are c ð1, 2Þ ¼
1 fað1Þbð2Þ bð1Það2Þg 21=2
ð7:40Þ
or, in more detail, c ðr 1 , r 2 Þ ¼
2
|–|
(a)
0
r1 – r2
|+|2
(b)
0
r1 – r2
Fig. 7.10 (a) The formation of a
Fermi hole by spin-correlation and (b) the formation of a Fermi heap when the spins are paired.
1 fc ðr 1 Þcn2 l2 ml2 ðr 2 Þ cn2 l2 ml2 ðr 1 Þcn1 l1 ml1 ðr 2 Þg 21=2 n1 l1 ml1
where the individual functions are hydrogenic atomic orbitals with Z ¼ 2. The striking feature of this result is that the degeneracy of the two product functions a(1)b(2) and b(1)a(2) is removed by the electron repulsion, and their two linear combinations c differ in energy by 2K. The exchange integral has no classical counterpart, and should be regarded as a quantum mechanical correction to the Coulomb integral J. However, despite its quantum mechanical origin, it is possible to discern the origin of this correction by considering the amplitudes c as one electron approaches the other. The crucial point is that c ¼ 0 when r1 ¼ r2 whereas cþ does not necessarily vanish. The corresponding differences in the probability densities are illustrated in Fig. 7.10. We see that there is zero probability of finding the two electrons in the same infinitesimal region of space if they are described by the wavefunction c , but there is no such restriction if their wavefunction is cþ (indeed, there is a small enhancement in the probability that they will be found together). The dip in the probability density jc j2 wherever r1 r2 is called a Fermi hole. It is a purely quantum mechanical phenomenon, and has nothing to do with the charge of the electrons; even ‘uncharged electrons’ would exhibit this phenomenon. It follows from the existence of the Fermi hole, that electrons that occupy c tend to avoid one another. Therefore, the average of the electron– electron repulsion energy can be expected to be lower for c than for cþ , for in the latter the electrons tend to be found near one another. The effect on the energy accounts for the reduction of the Coulombic potential energy from J to J K for electrons in c and its increase from J to J þ K for electrons in cþ .
224
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
It is appropriate at this point to mention something that will prove to be of crucial importance shortly. The wavefunction c is antisymmetric under the interchange of the names of the electrons: 1 fað2Þbð1Þ bð2Það1Þg 21=2 1 ¼ 1=2 fað1Þbð2Þ bð1Það2Þg ¼ c ð1, 2Þ 2
c ð2, 1Þ ¼
whereas cþ is symmetric under particle interchange: cþ ð2, 1Þ ¼ ¼
1 21=2 1 21=2
fað2Þbð1Þ þ bð2Það1Þg fað1Þbð2Þ þ bð1Það2Þg ¼ cþ ð1, 2Þ
7.10 The spectrum of helium At this stage we have seen that when both electrons are in the same orbital (as in 1s2, the ground state), the configuration gives rise to a single term with energy 2Ea þ J, with both Ea and J depending on the orbital that is occupied. When the two electrons occupy different orbitals (as in 1s12s1), then the configuration gives rise to two terms, one with energy Ea þ Eb þ J K and the other with energy Ea þ Eb þ J þ K. The separation of the terms by 2K should be detectable in the spectrum, and so we shall now consider the transitions in more detail. The ground-state configuration is 1s2. Its total orbital angular momentum is zero (because l1 ¼ l2 ¼ 0), so L ¼ 0 and it gives rise to an S term. The only excited configurations that we need consider in practice are those involving the excitation of a single electron, and therefore having the form 1s1nl1, because the excitation of two electrons exceeds the ionization energy of the atom. The configuration 1s1nl1 gives rise to terms with L ¼ l because only one of the electrons may have a non-zero orbital angular momentum. Therefore, the terms we have to consider are 1s12s1 S, 1s12p1 P, and so on. The selection rule Dl ¼ 1 implies that transitions may occur between S and P terms, between P and D, etc., but not between S and D. We need to consider the selection rules governing transitions between states of the form cþ and c described above. It turns out (as we demonstrate below) that the selection rules are symmetrical $ symmetrical
antisymmetrical $ antisymmetrical
but transitions between symmetrical and antisymmetrical combinations are not allowed. The basis of this selection rule is the vanishing of the transition dipole moment for states with different permutation symmetry. The electric dipole moment operator for a two-electron system is equal to er1 er2, which is symmetric under the permutation of the labels 1 and 2. The dipole moment for the transition between states of different permutation symmetry is Z þ ¼ e cþ ðr 1 , r 2 Þðr 1 þ r 2 Þc ðr 1 , r 2 Þdt1 dt2
7.11 THE PAULI PRINCIPLE
Orbitally Orbitally symmetric antisymmetric 1 3 1 1 3 S P D S P 3D 1
1s 3s
1
1s13d 1s13p1
1
1 s 3d 1
1
1
1s 3s 1
1
1s 2p 1s12s1 Allowed
Forbidden
1
1
1s 3p
1s 2p
1
1
1
Allowed 1 1 1s 2s
1s 2
Fig. 7.11 The energy levels of a
helium atom, their classification as singlets and triplets, and some of the allowed and forbidden transitions.
j
225
However, under the interchange of the labels 1 and 2, the integrand changes sign. As the value of an integral cannot depend on the labels that we give to the electrons, it follows that the only possible value for the integral is zero. Hence, there can be no transitions between symmetric and antisymmetric combinations. Finally, we need to consider the multiplicities of the terms. Because each electron has s ¼ 12, we expect S ¼ 0 and 1, corresponding to singlet and triplet terms, respectively. For the singlet terms, J ¼ L; for the triplet terms, the Clebsch–Gordan series gives J ¼ L þ 1, L, L 1 provided that L > 0. Thus, we can expect levels such as 1P1, 3P2, 3P1, and 3P0 to stem from each 1s1np1 configuration, and these levels are expected to be split by the spin–orbit coupling. At this stage (a phrase intended to strike a note of warning), we expect each of these terms to exist as the symmetric and antisymmetric combinations. So we expect eight terms to stem from a 1s1np1 configuration, with a symmetric and antisymmetric combination for each of 1P1, 3P2, 3P1, and 3P0. Similarly we expect (but see below) four terms from a 1s1ns1 configuration, corresponding to the symmetric and antisymmetric combinations for each of 1S and 3S. The observed spectrum of helium is, to some extent, consistent with these remarks. Each 1s1nl1 configuration gives rise to two types of term (Fig. 7.11), one symmetric and the other antisymmetric. We know which is which, because the ground-state configuration must be symmetric (both electrons occupy the same orbital), and therefore only symmetric states have appreciable transition intensity to the ground state. Furthermore, wherever both types of term can be identified, the antisymmetrical combination (the one that does not make transitions to the ground state) lies lower in energy than the symmetrical combination, in accord with the discussion in Section 7.9.7 There is, however, an extraordinary feature. An analysis of the spectrum shows that all the symmetric states are singlets and all the antisymmetric states are triplets. There are no symmetric triplets and no antisymmetric singlets. Moreover, there are only four terms from each 1s1np1 configuration, not eight. In fact, half of all possible terms appear to be excluded.
7.11 The Pauli principle The explanation of the omission of half the expected terms requires the introduction of an entirely new fundamental feature of nature. This was recognized by Wolfgang Pauli, who proposed the following solution. .......................................................................................................
7. Not too much should be made of this point. Although the analysis has shown that it is plausible that an antisymmetric combination, with its Fermi hole, should lie lower in energy, the conclusion was based on first-order perturbation theory and therefore ignored the distortion of the wavefunction that may occur. It turns out that this distortion, which corresponds to the shrinkage of the antisymmetric combination wavefunction so that the electrons lie closer to the nucleus than they do in the symmetric combination wavefunction, is of dominating importance for determining the order of energy levels. It remains true that the antisymmetric combination has a lower energy, but the reason is more complicated than the first-order argument suggests.
226
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
Consider the state of the system when the spins of the electrons are taken into account. In Section 4.12 we saw that the spin state of two electrons corresponding to S ¼ 0 is s ð1, 2Þ ¼
1 fað1Þbð2Þ bð1Það2Þg 21=2
where, as usual, a denotes the state with ms ¼ þ 12 and b denotes the state with ms ¼ 12. The state s is antisymmetric under particle exchange: s ð2, 1Þ ¼ s ð1, 2Þ On the other hand, the three states that correspond to S ¼ 1 are all symmetric under particle interchange: ðþ1Þ
sþ ð1, 2Þ ¼ að1Það2Þ 1 ð0Þ sþ ð1, 2Þ ¼ 1=2 fað1Þbð2Þ þ bð1Það2Þg 2 ð1Þ sþ ð1, 2Þ ¼ bð1Þbð2Þ (The superscript is the value of MS.) We can now list all combinations of orbital and spin states that might occur: c s ð0Þ
c sþ
c þ ð0Þ
c þ þ
ðþ1Þ
c þ þ
ð1Þ
c þ þ
c þ c þ
ðþ1Þ ð1Þ
The experimentally observed states have been printed with a tinted background. It is clear that there is a common feature: the allowed states are all antisymmetrical overall under particle interchange. This observation has been elevated to a general law of nature: The Pauli principle: The total wavefunction (including spin) must be antisymmetric with respect to the interchange of any pair of electrons. In fact, the Pauli principle can be expressed more broadly by recognizing that elementary particles can be classified as fermions or bosons. A fermion is a particle with half-integral spin; examples are electrons and protons. A boson is a particle with integral spin, including 0. Examples of bosons are photons (spin 1) and a-particles (helium-4 nuclei, spin 0). The more general form of the Pauli principle is then as follows: The total wavefunction must be antisymmetric under the interchange of any pair of identical fermions and symmetrical under the interchange of any pair of identical bosons. We shall consider only the restricted ‘electron’ form of the principle here, but use the full principle later (in Section 10.7). The principle should be regarded as one more fundamental postulate of quantum mechanics in addition to those presented in Chapter 1. However, it does have a deeper basis, for it can be rationalized to some extent by using relativistic arguments and the requirement that the total energy of the universe be positive. For us, it is a succinct, subtle, summary of experience (the spectrum of helium) that, as we shall see, has wide and never transgressed implications for the structure and properties of matter.
7.11 THE PAULI PRINCIPLE
j
227
It is a direct consequence of the Pauli principle that there is a restriction on the number of electrons that can occupy the same state. This implication of the Pauli principle is called the Pauli exclusion principle: No two electrons can occupy the same state. In its simplest form, the derivation of the exclusion principle from the Pauli principle runs as follows. Suppose the spin states of two electrons are the same. We can always choose the z-direction such that their joint spin state is a1a2, which is symmetric under particle interchange. According to the Pauli principle, the orbital part of the overall wavefunction must be antisymmetric, and hence of the form a(1)b(2) b(1)a(2). But if a and b are the same wavefunctions, then this combination is identically zero for all locations of the two electrons. Therefore, such a state does not exist, and we cannot have two electrons with the same spins in the same orbital. If the two electrons do not have the same spin, then there does not exist a direction where their joint spin state is a1a2, so the argument fails. It follows that if two electrons do occupy the same spatial orbital, then they must pair; that is, have opposed spins. (Note that ‘opposed spins’ does not mean that the spin part of the total wavefunction is a(1)b(2) or b(1)a(2) but rather the antisymmetric linear combination s (1,2).) Overall wavefunctions that satisfy the Pauli principle are often written as a Slater determinant. To see how such a determinant is constructed, consider another way of expressing the (overall antisymmetric) wavefunction of the ground state of helium: cð1, 2Þ ¼ c1s ðr 1 Þc1s ðr 2 Þs ð1, 2Þ 1 ¼ 1=2 c1s ðr 1 Þc1s ðr 2 Þfað1Þbð2Þ bð1Það2Þg 2 1 c1s ðr 1 Það1Þ c1s ðr 1 Þbð1Þ ¼ 1=2 2 c ðr Það2Þ c ðr Þbð2Þ 1s
The expansion of a 2 2 determinant is a b c d ¼ ad bc
2
1s
2
It is easy to show that the expansion of the determinant generates the preceding line. We now simplify the appearance of the Slater determinant by introducing the concept of a spinorbital, a joint spin–space state of the electron: ca1s ð1Þ ¼ c1s ðr 1 Það1Þ cb1s ð1Þ ¼ c1s ðr 1 Þbð1Þ Then the ground state can be expressed more succinctly as the following determinant: 1 ca1s ð1Þ cb1s ð1Þ cð1, 2Þ ¼ 1=2 a 2 c1s ð2Þ cb1s ð2Þ This is an example of a Slater determinant. The determinant displays the overall antisymmetry of the wavefunction very neatly, because if the labels 1 and 2 are interchanged, then the rows of the determinant are interchanged, and it is a general property of determinants that the interchange of two rows results in a change of sign.
228
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
Now suppose that the electrons have the same spin and occupy the same orbitals. The Slater determinant for such a state would be 1 ca ð1Þ ca1s ð1Þ cð1, 2Þ ¼ 1=2 1s a a c1s ð2Þ c1s ð2Þ 2 Because a determinant with two identical columns has the value 0 (another general property of determinants that can be easily verified in this case), this Slater determinant is identically zero. Such a state, therefore, does not exist, as required by the Pauli exclusion principle. The general form of a Slater determinant composed of the spinorbitals fa, fb, . . . and containing N electrons is fa ð1Þ fb ð1Þ fz ð1Þ 1=2 f ð2Þ f ð2Þ f ð2Þ b z 1 a ð7:41Þ cð1, 2, . . ., NÞ ¼ .. .. .. . N! . . f ðNÞ f ðNÞ f ðNÞ a b z A Slater determinant has N rows and N columns because there is one spinorbital for each of the N electrons present. The state is fully antisymmetric under the interchange of any pair of electrons, because that operation corresponds to the interchange of a pair of rows in the determinant. Furthermore, if any two spinorbitals are the same, then the determinant vanishes because it has two columns in common. Instead of writing out the determinant in full, which is tiresome, it is normally summarized by its principal diagonal: 1=2 1 cð1, 2, . . ., NÞ ¼ detjfa ð1Þfb ð2Þ . . . fz ðNÞj ð7:42Þ N! We are now in a position to return to the helium spectrum. We have seen that two electrons tend to avoid each other if they are described by an antisymmetric spatial wavefunction. However, if the two electrons are described by such a wavefunction, it follows that their spin state must be symmetrical, and hence correspond to S ¼ 1. Therefore, we can summarize the effect by saying that parallel spins tend to avoid one another. This effect is called spin correlation. However, the preceding discussion has shown that spin correlation is only an indirect consequence of spin working through the Pauli principle. That is, if the spins of the electrons are parallel, then the Pauli principle requires them to have an antisymmetric spatial wavefunction, which implies that the electrons cannot be found at the same point simultaneously. A consequence of spin correlation is, as we have seen, that the triplet term arising from a configuration lies lower in energy than the singlet term of the same configuration. The point should be noted, however, that the difference in energy is a similar indirect consequence of the relative spin orientations of the electrons and does not imply a direct interaction between spins. The difference in energy of terms of different multiplicity is a purely Coulombic effect that reflects the influence of spin correlation on the relative spatial distribution of the electrons.
7.12 PENETRATION AND SHIELDING
j
229
Many-electron atoms We have seen that a crude description of the ground state of the helium atom is 1s2 with both electrons in hydrogenic 1s-orbitals with Z ¼ 2. An improved description takes into account the repulsion between the electrons and the consequent swelling of the atom to minimize this disadvantageous contribution to the energy. It turns out that the effect of this repulsion on the orbitals occupied can be simulated to some extent by replacing the true nuclear charge, Ze, by an effective nuclear charge, Zeffe.8 The optimum value for helium, in the sense of corresponding to the lowest energy (recall the variation principle, Section 6.9), is Zeff 1.3. This approach to the description of atomic structure can be extended to other many-electron atoms, and we shall give a brief description of what is involved. Some of the principles will be familiar from elementary chemistry and we shall not dwell on them unduly.
7.12 Penetration and shielding Most descriptions of atomic structure are based on the orbital approximation, where it is supposed that each electron occupies its own atomic orbital, and that orbital bears a close resemblance to one of the hydrogenic orbitals. This is the justification of expressing the structure of an atom in terms of a configuration, such as 1s22s22p6 for neon. Thus, we write the wavefunction for the neon atom in the orbital approximation as 1=2 1 det 1sa ð1Þ1sb ð2Þ . . . 2pb ð10Þ c¼ 10! It must clearly be understood that this expression is an approximation, because the actual many-electron wavefunction is not a simple product (or a sum of such simple products) but is a more general function of 3N variables and two spin states for each electron. To reproduce the exact wavefunction, we would have to take a superposition of an infinite number of antisymmetric products, as discussed in Chapter 9. According to the Pauli exclusion principle, a maximum of two electrons can occupy any one atomic orbital. As a result, the electronic structure of an atom consists of a series of concentric shells of electron density, where a shell consists of all the orbitals of a given value of n. We refer to the K-shell for n ¼ 1, the L-shell for n ¼ 2, the M-shell for n ¼ 3, and so on. The Li atom (Z ¼ 3), for example, consists of a complete K-shell and one electron in one of the orbitals of the L-shell. Each shell consists of n subshells, which are the orbitals with a common value of l. There are 2l þ 1 individual orbitals in a subshell. In a hydrogenic atom, all subshells of a given shell are degenerate, but the presence of electron–electron interactions in many-electron atoms removes this degeneracy, and although the members of a given subshell remain degenerate (so the three 2p-orbitals are degenerate in all atoms), the .......................................................................................................
8. In casual usage, Zeff itself rather than Zeffe is commonly called the ‘effective nuclear charge’.
230
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
subshells correspond to different energies. It is typically found, for valence (outermost) electrons at least, that the energies of the subshells lie in the order s < p < d < f, but there are deviations from this simple rule. The explanation of the order of subshells is based on the central-field approximation, in which the highly complicated inter-electronic contribution to the energy, which for electron 1 is V¼
X i6¼1
e2 4pe0 r1i
ð7:43Þ
is replaced by a single point negative charge on the nucleus, so V
se2 4pe0 r1
ð7:44Þ
where se is an effective charge that repels the charge e of the electron of interest. As a result of this approximation, the nuclear charge Ze is reduced to (Z s)e, and hence we can write Zeff ¼ Z s
+Ze
–e
Fig. 7.12 According to classical
electrostatics, the charge of a spherically symmetrical distribution can be represented by a point charge equal in value to the total charge of the region and placed at its centre.
ð7:45Þ
The quantity s is called the nuclear screening constant and is characteristic of the orbital that the electron (which we are calling 1) occupies. Thus, s is different for 2s- and 2p-orbitals. It also depends on the configuration of the atom, and s for a given orbital has different values in the ground and excited states. The partial justification for this seemingly (and actually) drastic approximation comes from classical electrostatics. According to classical electrostatics, when an electron is outside a spherical region of electric charge, the potential it experiences is the same as that generated by a single point charge at the centre of the region with a magnitude equal to the total charge within a sphere that cuts through the position of the electron (Fig. 7.12). Thus, if the K-shell is full and very compact, the effect of its two electrons can be simulated by placing a point charge 2e on the nucleus provided that the electron of interest stays wholly outside the core region (here the region of the K-shell electrons) of the atom. If the electron of interest wanders into the core, then its interaction varies the closer it is to the nucleus, and when it is at the nucleus, it experiences the full nuclear charge. The reduction of the nuclear charge due to the presence of the other electrons in an atom is called shielding, and its magnitude is determined by the extent of penetration of core regions of the atom, the extent to which the electron of interest will be found close to the nucleus and inside spherical shells of charge. Strictly speaking, the shielding constant varies with distance, and an electron does not have a single value of s. However, in the next approximation we replace the varying value of s by its average value, and hence treat Zeff as a constant typical of the atom and of the orbital occupied by the electron of interest. This is the basis for replacing Z ¼ 2 by Z ¼ 1.3 for each electron in a He atom, for we are ascribing the average value s ¼ 0.7 to each electron. It follows from the discussion of the radial distribution functions for electrons in atoms (Section 3.12) that an ns-electron penetrates closer
7.13 PERIODICITY 2
2
Energy
S
3s
2
P
1 2
S
3p1 2P 2p
2
D
F
3d 1 2D
1 2
2s
P
1 2
S
Fig. 7.13 A schematic indication of
the orbital energy levels of a manyelectron atom (lithium, in fact) showing the removal of the degeneracy characteristic of hydrogenic atoms.
1s 2p
2s
3p
3s 4s
3d
5s
4d
4p 5p
6s
5d
7s
6d
6p
4f 5f
Fig. 7.14 The order of occupation of
energy levels as envisaged in the building-up principle. At the end of each period, revert to the start of the next period.
j
231
to the nucleus than does an np-electron. Hence we can expect an ns-electron to be less shielded by the core electrons than an np-electron, and hence to have a lower energy. There is a similar difference between np- and nd-electrons, for the wavefunctions of the latter are proportional to r2 whereas those of the former are proportional to r close to the nucleus; hence, nd-electrons are excluded more strongly from the nucleus than npelectrons. These effects can be seen in the atomic energy-level diagram for Li (Fig. 7.13), which has been inferred from an analysis of its emission spectrum.
7.13 Periodicity The ground-state electron configurations of atoms are determined experimentally by an analysis of their spectra or, in some cases, by magnetic measurements. These configurations show a periodicity that mirrors the block, group, and period structure of the periodic table. The rationalization of the observed configurations is normally expressed in terms of the building-up (or aufbau) principle. According to this principle, electrons are allowed to occupy atomic orbitals in an order that mirrors the structure of the periodic table (Fig. 7.14) and subject to the Pauli exclusion principle that no more than two electrons can occupy any one orbital, and if two electrons do occupy an orbital, then their spins must be paired. The order of occupation largely follows the order of energy levels as determined by penetration and shielding, with ns-orbitals being occupied before np-orbitals. The lowering of the energy of ns-orbitals is so great that in certain regions of the table they lie below the (n 1)d-orbitals of an inner shell: the occupation of 4s-orbitals before 3d-orbitals is a well-known example of this phenomenon, and it accounts for the intrusion of the d-block into the structure of the periodic table. It is too much to expect such a simple procedure based on the energies of one-electron orbitals to account for all the subtleties of the periodic table. What matters is the attainment of the lowest total energy of the atom, not the lowest sum of one-electron energies, for the latter largely ignores electron– electron interactions (except implicitly). Thus, it is found in some cases that the lowest total energy of the atom is attained by shipping electrons around: the favouring of d5 and d10 configurations is an example of a manner in which the atom can relocate electrons to minimize the total energy, perhaps at the expense of having to occupy an orbital of higher energy. There are various regions of the periodic table where it is necessary to adjust the configuration suggested by the building-up principle, but it is a remarkably simple and generally reliable principle for accounting for the subtleties of the properties of atoms. There are two features of the building-up principle that should be kept in mind. One is that when more than one orbital is available for occupation, electrons occupy separate orbitals before entering an already half-occupied orbital. This gives them a greater spatial separation, and hence minimizes the total energy of the atom. Second, when electrons occupy separate orbitals, they do so with parallel spins. This rule is often called Hund’s rule of
232
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
maximum multiplicity (Section 7.17), and it can be traced to the effects of spin correlation, as we have already seen for helium. Example 7.4 The ground-state electron configurations of atoms
What is the ground-state electron configuration of carbon? Method. This example is a recapitulation of material normally encountered in
introductory chemistry courses, but is included to illustrate the foregoing material. Decide on the number of electrons in the atom, then let them occupy the available orbitals in accord with the scheme in Fig. 7.14 and the restrictions of the Pauli principle. When dealing with the outermost electrons, allow electrons to occupy separate degenerate orbitals, and take note of Hund’s rule. Answer. The six electrons of carbon (Z ¼ 6) have a ground-state configuration 1s22s22p2. In more detail, we expect the configuration 1s2 2s2 2p1x 2p1y , with the two 2p-electrons having parallel spins. A triplet is therefore expected for the ground term of the atom. Comment. It should always be remembered that an electron configuration has
meaning only within the orbital approximation. Self-test 7.4. What is the ground-state electron configuration of an O atom? [1s2 2s2 2p2x 2p1y 2p1z , a triplet term]
Once the ground-state electron configuration of an atom is known, it is possible to go on to rationalize a number of the properties. For example, the ionization energy, I, of an element, the minimum energy needed to remove an electron from a gas-phase (ground-state) atom of an element at T ¼ 0 (that is, absolute zero)
14.53 13.61 17.42 5.14
5.39
9.32 8.30 11.26
Ionization energy, I /eV 13.60
21.56
24.58
EðgÞ ! Eþ ðgÞ þ e ðgÞ
H He Li Be B C N O F NeNa Fig. 7.15 The variation of the first
ionization energy through Period 2 of the periodic table.
generally increases across a period because the effect of nuclear attraction on the outermost electron increases more rapidly than the repulsion from the additional electrons that are present. However, the variation is not uniform (Fig. 7.15), because account must be taken of the identity of the orbitals from which the outermost electron is removed and the energy of the ion remaining after the loss of the electron. The dip between Be and B, for instance, can be explained on the grounds that their electron configurations are 1s22s2 and 1s22s22p1, respectively; so ionization takes place from a 2p orbital in B but a 2s orbital in Be, and the latter orbital has a lower energy on account of its shielding and penetration. The decrease between N and O reflects the fact that in N (1s2 2s2 2p1x 2p1y 2p1z ) the electron is removed from a half-filled 2porbital, whereas in O (1s2 2s2 2p2x 2p1y 2p1z ) the electron is ‘helped’ on its way by another electron that is present in the 2px-orbital and the fact that the resulting 2p3 configuration (being a half-filled configuration) has a low energy. The steep fall in ionization energy between He and Li, and between Ne and Na, reflects the fact that the electron is being removed from a new shell and so is more distant from the nucleus.
7.14 SLATER ATOMIC ORBITALS
j
233
Table 7.1 Values of Zeff ¼ Z s for neutral ground-state atoms
H 1s
He
1 Li
1.6875 Be
B
C
N
O
F
Ne
1s
2.6906
3.6848
4.6795
5.6727
6.6651
7.6579
8.6501
9.6421
2s
1.2792
1.9120
2.5762
3.2166
3.8474
4.4916
5.1276
5.7584
2.4214
3.1358
3.8340
4.4532
5.1000
5.7584
2p Na
Mg
Al
Si
P
S
Cl
Ar
1s
10.6259
11.6089
12.5910
13.5754
14.5578
15.5409
16.5239
17.5075
2s
6.5714
7.3920
8.2136
9.0200
9.8250
10.6288
11.4304
12.2304
2p
6.8018
7.8258
8.9634
9.9450
10.9612
11.9770
12.9932
14.0082
3s
2.5074
3.3075
4.1172
4.9032
5.6418
6.3669
7.0683
7.7568
4.0656
4.2852
4.8864
5.4819
6.1161
6.7641
3p
Values are from E. Clementi and D.L. Raimondi, Atomic screening constants from SCF functions. IBM Res. Note NJ-27 (1963).
7.14 Slater atomic orbitals No definitive analytical form can be given for the atomic orbitals of manyelectron atoms because the orbital approximation is very primitive. Nevertheless, it is often helpful to have available a set of approximate atomic orbitals which model the actual wavefunctions found by using the more sophisticated numerical techniques that we describe in Chapter 9. These Slater type orbitals (STOs) are constructed as follows: 1. An orbital with quantum numbers n, l, and ml belonging to a nucleus of an atom of atomic number Z is written cnlml ðr, y, fÞ ¼ Nrneff 1 eZeff r=neff Ylml ðy, fÞ where N is a normalization constant, Ylml is a spherical harmonic (Table 3.1), and r ¼ r/a0. 2. The effective principal quantum number, neff, is related to the true principal quantum number, n, by the following mapping: n ! neff :
1!1 2!2 3!3
4 ! 3:7 5 ! 4:0 6 ! 4:2
3. The effective nuclear charge, Zeff, is taken from Table 7.1. The values in Table 7.1 have been constructed by fitting STOs to numerically computed wavefunctions,9 and they supersede the values that were originally given by Slater in terms of a set of rules. Care should be taken when using STOs because orbitals with different values of n but the .......................................................................................................
9. The procedure by which the value of Zeff ¼ 1.6875 for He shown in Table 7.1 was obtained differs from the procedure based on the variation principle which yields the value of 1.3 quoted in Section 7.12.
234
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
same values of l and ml are not orthogonal to one another. Another deficiency of STOs is that ns-orbitals with n > 1 have zero amplitude at the nucleus.
7.15 Self-consistent fields The best atomic orbitals are found by numerical solution of the Schro¨dinger equation. The original procedure was introduced by Hartree and is known as the self-consistent field (SCF) method. The procedure was improved by Fock and Slater to include the effects of electron exchange, and the orbitals obtained by their methods are called Hartree–Fock orbitals.10 The assumption behind the technique is that any one electron moves in a potential which is a spherical average of the potential due to all the other electrons and the nucleus, and which can be expressed as a single charge centred on the nucleus (this is the central-field approximation; but it is not assumed that the charge has a fixed value). Then the Schro¨dinger equation, a differential equation, is integrated numerically for that electron and that spherically averaged potential, taking into account the fact that the total charge inside the sphere defined by the position of the electron varies as the distance of the electron from the nucleus varies (recall Fig. 7.12). This approach supposes that the wavefunctions of all the other electrons are already known so that the spherically averaged potential can be calculated. That is not in general true, so the calculation starts out from some approximate form of the wavefunctions, such as approximating them by STOs. The Schro¨dinger equation for the electron is then solved, and the procedure is repeated for all the electrons in the atom. At the end of this first round of calculation, we have a set of improved wavefunctions for all the electrons. These improved wavefunctions are then used to calculate the spherically averaged potential, and the cycle of computation is performed again. The cycle is repeated until the improved set of wavefunctions does not differ significantly from the wavefunctions at the start of the cycle. The wavefunctions are then self-consistent, and are accepted as good approximations to the true many-electron wavefunction. The Hartree–Fock equations on which the procedure is based are slightly tricky to derive (see Further information 11) but they are reasonably easy to interpret. The hamiltonian that we need to consider is H¼
X i
hi þ 12
X i; j
0
e2 4pe0 rij
ð7:46Þ
where hi is a hydrogenic hamiltonian for electron i in the field of a bare nucleus of charge Ze. This operator is called the core hamiltonian. The factor of 12 in the double sum prevents the double-counting of interactions. The prime on the summation excludes terms for which i ¼ j as electrons do .......................................................................................................
10. See Chapter 9 for a more detailed account of the Hartee–Fock procedure.
7.15 SELF-CONSISTENT FIELDS
j
235
not interact with themselves. The Hartree–Fock equation for a space orbital (spatial wavefunction) cs occupied by electron 1 is ( ) X h1 þ ð2Jr Kr Þ cs ð1Þ ¼ es cs ð1Þ ð7:47Þ r
The sum is over all occupied spatial wavefunctions. The terms Jr and Kr are operators that have the following effects. The Coulomb operator, Jr, is defined as follows: Z
e2 c ð2Þdt2 cs ð1Þ cr ð2Þ ð7:48Þ Jr cs ð1Þ ¼ 4pe0 r12 r This operator represents the Coulombic interaction of electron 1 with electron 2 in the orbital cr. The exchange operator, Kr, is defined similarly: Z
e2 c ð2Þdt2 cr ð1Þ cr ð2Þ ð7:49Þ Kr cs ð1Þ ¼ 4pe0 r12 s This operator takes into account the effects of spin correlation. The quantity es in eqn 7.47 is the one-electron orbital energy. Equations 7.48 and 7.49 show that it is necessary to know all the other spatial wavefunctions in order to set up the operators J and K and hence to find the form of each wavefunction. Once the final, self-consistent form of the orbitals has been established, we can find the orbital energies by multiplying both sides of eqn 7.47 by cs (1) and integrating over all space. The right-hand side is simply es, and so Z X es ¼ cs ð1Þh1 cs ð1Þdt1 þ ð2Jsr Ksr Þ ð7:50Þ r
where
Z
cs ð1ÞJr cs ð1Þdt1 Z e2 1 c ð2Þcs ð1Þdt1 dt2 cs ð1Þcr ð2Þ ¼ r12 r 4pe0
Jsr ¼
ð7:51Þ
which, after reorganizing the integrand a little, is seen to be the Coulomb integral introduced in connection with the structure of helium (eqn 7.36). It is the average potential energy of interaction between an electron in cs and an electron in cr. Similarly, Z Ksr ¼ cs ð1ÞKr cs ð1Þdt1 Z e2 1 c ð2Þcr ð1Þdt1 dt2 cs ð1Þcr ð2Þ ¼ ð7:52Þ r12 s 4pe0 This integral is recognizable, after some reorganization, as the exchange integral (eqn 7.37). In passing, note that Krr ¼ Jrr. The sum of the orbital energies is not the total energy of the atom, for such a sum counts all electron–electron interactions twice. So, to obtain the total energy we need to eliminate the effects of double counting: X X E¼2 es ð2Jrs Krs Þ ð7:53Þ s
r; s
236
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
where the sum is over the occupied orbitals (each of which is doubly occupied in a closed-shell species). We can verify that this procedure gives the correct result for helium in the ground-state configuration 1s2. The one-electron energy is e1s ¼ E1s þ ð2J1s;1s K1s;1s Þ ¼ E1s þ J1s;1s and the total energy is E ¼ 2e1s ð2J1s;1s K1s;1s Þ ¼ 2ðE1s þ J1s, 1s Þ J1s;1s ¼ 2E1s þ J1s;1s exactly as before. The energy required to remove an electron from an orbital cr, on the assumption that the remaining electrons do not adjust their distributions, is the one-electron energy er. Therefore, we may equate the one-electron orbital energy with the ionization energy of the electron from that orbital. This identification is the content of Koopmans’ theorem:
Electron density
Ir er
K L M Distance from nucleus, r
Fig. 7.16 A representation of the
electron density calculated for a many-electron atom. Note that the shell structure is apparent, but that the total electron density falls to zero monotonically. The graph is a plot of the electron density along a radius, not the radial distribution function.
ð7:54Þ
The theorem is only an approximation, because the remaining N 1 electrons have a different set of Hartree–Fock orbital energies (in the N 1 electron ion) than they did in the N electron atom. (The spherically averaged electrostatic potentials differ in the N and N 1 electron systems.) Solutions of the Hartree–Fock equations are generally given either as numerical tables or fitted to sets of simple functions. Once they are available, the total electron density in an atom may be calculated very simply by summing the squares of the wavefunctions for each electron. As Fig. 7.16 shows, the calculated value exhibits the shell structure of the atom that more primitive theories have led us to expect. Note that the total electron density shows the shell structure as a series of inflections: it decreases monotonically without intermediate maxima and minima. Hartree–Fock SCF atomic orbitals are by no means the most refined orbitals that can be obtained. They are rooted in the orbital approximation and therefore to an approximate central-field form of the potential. The true wavefunction for an atom, whatever that may be, depends explicitly on the separations of the electrons, not merely their distances from the nucleus. The incorporation of the separations rij explicitly into the wavefunction is the background of the correlation problem, which is at the centre of much modern work (Chapter 9). Another route to improvement is to use the Dirac equation for the calculation rather than the non-relativistic Schro¨dinger equation. Relativistic effects are of considerable importance for heavy atoms, and are needed to account for various properties of the elements, including the colour of gold, the lanthanide contraction, the inert-pair effect, and even the liquid character of mercury.
7.16 Term symbols and transitions of many-electron atoms The state of a many-electron atom is expressed by a term symbol of exactly the same kind as we have already described (Section 7.6). To construct the symbol, we need to know the total spin, S, the total orbital angular momentum, L,
7.16 TERM SYMBOLS AND TRANSITIONS OF MANY-ELECTRON ATOMS
j
237
and the total angular momentum, J, of the atom.11 These quantities are constructed by an appropriate application of the Clebsch–Gordan series (Section 4.10). For instance, in the Russell–Saunders coupling scheme, the total angular momenta of the valence electrons are constructed as follows: S ¼ s1 þ s2 , s1 þ s2 1, . . ., js1 s2 j L ¼ l1 þ l2 , l1 þ l2 1, . . ., jl1 l2 j J ¼ L þ S, L þ S 1, . . ., jL Sj Each of these series may need to be applied several times if there are more than two electrons in the valence shell. The core electrons can be neglected because the angular momentum of a closed shell is zero. Example 7.5 The construction of term symbols
Construct the term symbols that can arise from the configurations (a) 2p13p1 and (b) 2p5. Method. First, construct the possible values of L by using the Clebsch–Gordan
series and identify the corresponding letters for the terms. Then construct the possible values of S similarly, and work out the multiplicities. Finally, construct the values of J from the values of L and S for each term by using the Clebsch–Gordan series again. A useful trick for shells that are more than half full is to consider the holes in the shell as particles, and to construct the term symbol for the holes. That is equivalent to treating the electrons, because a closed shell has zero angular momentum, and the angular momentum of the electrons must be equal to (in the sense of cancelling) the angular momentum of the holes. Answer. (a) For this configuration l1 ¼ 1 and l2 ¼ 1, so L ¼ 2, 1, 0, and the configuration gives rise to D, P, and S terms. Two electrons result in S ¼ 1, 0, giving rise to triplet and singlet terms, respectively, so the complete set of terms is 3D, 1D, 3P, 1P, 3S, and 1S. The values of J that can arise are formed from J ¼ L þ S, L þ S 1, . . . , jL Sj, and so the complete list of term symbols is 3
D3 , 3 D2 , 3 D1 , 1 D2 , 3 P2 , 3 P1 , 3 P0 , 1 P1 , 3 S1 , 1 S0
(b) The configuration 2p5 is equivalent to a single hole in a shell, so L ¼ l ¼ 1, corresponding to a P term. Because S ¼ s ¼ 12 for the hole, the term symbol is 2P. The two levels of this terms are 2P3/2 and 2P1/2, the same terms that arise from 2p1. Comment. The configuration 2p2 does not give rise to all the terms that 2p13p1
generates because the Pauli principle forbids the occurrence of certain combinations of spin and orbital angular momentum. This point is taken up below. Self-test 7.5. Construct the term symbols that can arise from the configura-
tions (a) 3d14p1 and (b) 3d9.
.......................................................................................................
11. The ‘total’ angular momentum J we are considering here takes into account contributions from electrons; angular momentum due to, for instance, nuclear spin is not being considered.
238
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
As indicated in the Comment in the example, some care is needed when deriving the term symbols arising from configurations of equivalent electrons, as in the 2p2 configuration of carbon. For instance, although the configuration gives rise to D, P, and S terms, and S ¼ 0, 1, it is easy to see that 3D is excluded. For this L ¼ 2 term to occur, we need to obtain a state with ML ¼ þ 2, as well as the other ML states that belong to a D term. To obtain ML ¼ þ 2, both electrons must occupy orbitals with ml ¼ þ1. However, because the two electrons are in the same orbital (n ¼ 2, l ¼ 1, ml ¼ þ1), they cannot have the same spins, so the S ¼ 1 state is excluded. A quick way to decide which combinations of L and S are allowed is to use group theory and to identify the antisymmetrized direct product (Section 5.14). For the 2p2 configuration we need to form Gð1Þ Gð1Þ ¼ Gð2Þ þ ½Gð1Þ þ Gð0Þ
ð7:55aÞ
Gð1=2Þ Gð1=2Þ ¼ Gð1Þ þ ½Gð0Þ
ð7:55bÞ
where we have used the notation introduced in Sections 5.14 and 5.20. (For eqn 7.55b, it should be apparent from Section 7.11 that [G(0)] is associated with the singlet spin state s which is antisymmetric under electron interchange. For eqn 7.55a, inspection of the vector coupling coefficients in Appendix 2 readily shows that [G(1)] is antisymmetric with respect to electron exchange.) To ensure that the overall state is antisymmetric, we need to associate symmetric with antisymmetric combinations. In this case the terms are 1D, 3P, and 1S. A more pedestrian procedure is to draw up a table of microstates, or combinations of orbital and spin angular momentum of each individual electron, and then to identify the values of L and S to which they belong. We shall denote the spinorbital as ml if the spin is a and ml if the spin is b. Then one typical microstate of two electrons would be ð1, 1Þ if one electron occupies an orbital with ml ¼ þ1 with a spin and the second electron occupies the same orbital with b spin. This microstate has ML ¼ þ 2 and MS ¼ 0, and is put into the appropriate cell in Table 7.2. The complete set of microstates can be compiled in this way, and ascribed to the appropriate cells in the table. Note that microstates such as (1,1), which correspond to the two a-spins in an
Table 7.2 The microstates of p2 ML, MS:
þ1
þ2 þ1
(1,0)
0
(1,1)
1
(1,0)
2
0
1
) (1,1 ), (1 ,0) (1,0 ), (1 ,1), (0,0 ) (1,1 (1,0), (1,0)
,0 ) (1 ,1 ) (1
) (1,1
,0 ) (1
7.17 HUND’S RULES AND THE RELATIVE ENERGIES OF TERMS
j
239
orbital with ml ¼ þ1, are excluded by the Pauli principle and have been omitted. Now we analyse the microstates to see to which values of L and S they belong. The microstate ð1, 1Þ must belong to L ¼ 2, S ¼ 0, which identifies it as a state of a 1D term. There are five states with L ¼ 2, so we can strike out one microstate in each row of the column headed MS ¼ 0; which one we strike out in each case is immaterial as this is only a bookkeeping exercise, and striking out one state is equivalent to striking out one possible linear combination. The next row shows that there is a microstate with ML ¼ þ1 and MS ¼ þ1. This state must belong to L ¼ 1 and S ¼ 1 and hence to the term 3P. The nine states of this term span ML ¼ þ1, 0, 1 and MS ¼ þ1, 0, 1, and so we can strike out nine of the remaining ten microstates. Only one microstate remains: it has ML ¼ 0 and MS ¼ 0, and hence belongs to L ¼ 0 and S ¼ 0, and is therefore a 1S term. Now we have accounted for all the microstates, and have identified the terms as 1D, 3P, and 1 S, as we had anticipated. The transitions that are allowed by the selection rules for a many-electron atom are DJ ¼ 0, 1 but J ¼ 0 ! J ¼ 0 forbidden DL ¼ 0, 1 Dl ¼ 1 DS ¼ 0 The rules regarding DJ and DL express the general point about the conservation of angular momentum. The rule concerning Dl is based on the conservation of angular momentum for the actual electron that is excited in the transition and its acquisition of the angular momentum of the photon; it is relevant when using a single Slater determinant to represent a state. The rule regarding DS reflects the fact that the electric component of the electromagnetic field can have no effect on the spin angular momentum of the electron, and in particular that it cannot induce transitions between wavefunctions that have different permutation symmetry (see Section 7.10). The selection rules on J are exact: those concerning l, L, and S presume that these individual angular momenta are well-defined.
7.17 Hund’s rules and the relative energies of terms Friedrich Hund devised a set of rules for identifying the lowest energy term of a configuration with the minimum of calculation. 1. The term with the maximum multiplicity lies lowest in energy. For the configuration 2p2, we expect the 3P term to lie lowest in energy. The explanation of the rule can be traced to the effects of spin correlation. On account of the existence of a Fermi hole, orbitals containing electrons with the same spin can contract towards the nucleus without an undue increase in electron–electron repulsion. The Fermi hole acts as a kind of protective halo around the electrons. 2. For a given multiplicity, the term with the highest value of L lies lowest in energy.
j
240
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
For example, if we had to choose between 3P and 3F in a particular configuration, then we would select the latter as the lower energy term. The classical basis of this rule is essentially that if electrons are orbiting in the same direction (and so have a high value of L), then they will meet less often than when they are orbiting in opposite directions (and so have a low value of L). Because they meet less often, their repulsion is less. 3. For atoms with less than half-filled shells, the level with the lowest value of J lies lowest in energy.
1
D
1
D
3p 2
1/r ij
1
D
3s 3d 1
1 3
3
D
D
1
D
Fig. 7.17 The effect of configuration
interaction between two D terms of the same multiplicity and the consequent reversal of the order of the terms of a configuration predicted by Hund’s rules.
For the 2p2 configuration, which corresponds to a shell that is less than half full, the ground term is 3P. It has three levels, with J ¼ 2, 1, 0. We therefore predict that the lowest energy level is 3P0. When a shell is more than half full, the opposite rule applies (highest J lies lowest in energy). The origin of this rule, in both its forms, is the spin–orbit coupling, and was discussed in Section 7.5. The rules are reasonably reliable for predicting the term of lowest energy, but are not particularly reliable for ranking all the terms according to their energy. One reason for their failure may be that the structure of an atom is inaccurately described by a single configuration, and a better description is in terms of configuration interaction, a superposition of several configurations. An example is found among the excited states of magnesium, and in particular the configuration 3s13d1, which is expected to have 3D energetically below 1D whereas the opposite is found to be the case. An explanation is that the 1D term is actually a mixture of about 75 per cent 3s13d1 and 25 per cent 3p2 (which can also give rise to a 1D term). If the two configurations have a similar energy, the electron–electron repulsion term perturbs them and, as for any two-level system, they move apart in energy (Fig. 7.17). The lower combination is pressed down in energy, and as a result may fall below the 3D term, which is unchanged because there is no 3p2 3D term. Configuration interaction is discussed in more detail in Chapter 9.
7.18 Alternative coupling schemes J L
l1
l2
S
s 1, s 2 Fig. 7.18 A vector representation of
Russell–Saunders (LS) coupling in a two-electron atom.
We have just seen that a configuration should not be taken too literally; the same is true of term symbols as well. The specification of a term symbol implies that L and S have definite values, but that may not be true when spin– orbit coupling is appreciable, particularly in heavy atoms. The term symbols we have introduced are based on Russell–Saunders (LS) coupling, which is applicable when spin–orbit coupling is weak in comparison with Coulombic interactions between electrons. When the latter are dominant, they result in the coupling of orbital angular momenta into a resultant with quantum number L and the spin angular momenta into a resultant with quantum number S. The weak spin–orbit interaction finally couples these composite angular momenta together into J (Fig. 7.18). To represent the relative strengths of the coupling of the angular momenta, we imagine the component vectors as precessing, or migrating around their cones, at a rate proportional to the strength of the coupling. When Russell– Saunders coupling is appropriate, the individual orbital momenta and the
7.18 ALTERNATIVE COUPLING SCHEMES
j1
s1
241
spin momenta precess rapidly around their resultants, but the two resultants L and S precess only slowly around their resultant, J. When spin–orbit coupling is strong, as it is in heavy atoms, we use jj-coupling. In this scheme, the orbital and spin angular momenta of individual electrons couple to give a combined angular momentum j, and then these combined angular momenta couple to give a total angular momentum J. Now l and s each precess rapidly around their resultant j, and the various js precess slowly around their resultant J (Fig. 7.19). In this scheme, L and S are not specified and so the term symbol loses its significance.
J
l1
j
j2
l2 s2
Example 7.6 Using the jj-coupling scheme
Fig. 7.19 A vector representation of
jj-coupling in a two-electron atom.
Within the jj-coupling scheme, what values of the total angular momentum J are permitted for the electron configuration 5p15d1? Method. Coupling l1 ¼ 1 and s1 ¼ 12 yields values for the resultant j1. Similarly,
coupling l2 ¼ 2 and s2 ¼ 12 yields values for the resultant j2. Then couple j1 and j2 to obtain J. Use the Clebsch–Gordan series in each case. Answer. Coupling l1 ¼ 1 and s1 ¼ 12 yields j1 ¼ 32, 12. Coupling l2 ¼ 2 and s2 ¼ 12
yields j2 ¼ 52, 32. Finally, coupling j1 and j2 gives values of J ¼ 4, 3, 3, 3, 2, 2, 2, 2, 1, 1, 1, 0. Self-test 7.6. Within the jj-coupling scheme, what values of the total angular
momentum J are permitted for the electron configuration 5d16d1?
Pure LS coupling
Pure jj coupling
1
S0
( 32 , 32 )
1
D2
( 32 , 12 )
Although the significance of the term symbol is lost when jj-coupling is relevant, symbols can still be used to label the terms because there is a correlation between Russell–Saunders and jj-coupled terms. To see that this is so, consider the np2 configurations of the Group 14 elements carbon to lead. In the Russell–Saunders scheme we expect 1S, 3P, and 1D terms, and the levels 3 P2, 3P1, and 3P0 of the 3P term. The energies of these terms are indicated on the left of Fig. 7.20. On the other hand, in jj-coupling, each p-electron can have either j ¼ 12 or j ¼ 32. The resulting total angular momenta will be Gð1=2Þ Gð1=2Þ ¼ Gð1Þ þ Gð0Þ Gð1=2Þ Gð3=2Þ ¼ Gð2Þ þ Gð1Þ
3
P2 P1 3 P0 3
Gð3=2Þ Gð3=2Þ ¼ Gð3Þ þ Gð2Þ þ Gð1Þ þ Gð0Þ ( 12 , 12 )
C Si Ge
Sn
Pb
Fig. 7.20 The correlation diagram
for a p2 configuration and the approximate location of Group 14 atoms. Note that Russell–Saunders terms can be used to label the atoms regardless of the extent of jj-coupling.
Because an electron with j ¼ 12 can be expected to have a lower energy than one with j ¼ 32 on the basis of spin–orbit coupling in a less than half-filled shell, we expect the order of energies indicated on the right of the illustration. Note that the Pauli principle excludes J ¼ 3, because to achieve it, both electrons would need to occupy the same orbital with the same spin. The states on the two sides can be correlated because J is well-defined in both coupling schemes and we know (Section 6.1) that states of the same symmetry (in this case, the same J) do not cross when perturbations are present. The resulting correlation of states is shown in the illustration, which
242
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
is called a correlation diagram. As can be seen, even though Russell–Saunders coupling is inappropriate for the heavier members of the group, it can still be used to construct labels for the terms.
Atoms in external fields In this final section, we shall consider how the application of electric and magnetic fields can affect the energy levels and hence the spectra of atoms. We shall describe two effects: the Zeeman effect is the response to a magnetic field; the Stark effect is the response to an electric field.
7.19 The normal Zeeman effect ML +1 1
P
0 –1
∆ML = –1
0
Electrons possess magnetic moments as a result of their orbital and spin angular momenta. These moments will interact with an externally applied magnetic field, and the resulting shifts in energy should be apparent in the spectrum of the atom. Consider first the effect of a magnetic field on a singlet term, such as 1P. Because S ¼ 0, the magnetic moment of the atom arises solely from the orbital angular momentum. For a field of magnitude b in the z-direction the hamiltonian is Hð1Þ ¼ mz b ¼ ge lz b
+1
ð7:56Þ
If several electrons are present, Hð1Þ ¼ Mz b ¼ ge ðlz1 þ lz2 þ Þb ¼ ge Lz b
1
ð7:57Þ
The first-order correction to the energy of the P term is therefore S
0
hb ¼ mB ML b Eð1Þ ¼ h1 PML jHð1Þ j1 PML i ¼ ge ML 1
B=0
B>0
~
Fig. 7.21 The splitting of energy
levels of an atom in the normal Zeeman effect, and the splitting of the transitions into three groups of coincident lines.
ð7:58Þ
where mB is the Bohr magneton (eqn 7.12). A S term has neither orbital nor spin angular momentum, so it is unaffected by a magnetic field. It follows that the transition 1P ! 1S should be split into three lines (Fig. 7.21), with a splitting of magnitude mB b. A 1 T magnetic field splits lines by only 0.5 cm1, so the effect is very small. This splitting of a spectral line into three components is an example of the normal Zeeman effect. The three transitions that make up 1P ! 1S correspond to different values of DML. We have already seen that transitions with different values of Dml (and likewise DML) correspond to different polarization of electromagnetic radiation. In the present case, an observer perpendicular to the magnetic field sees that the outer lines of the trio (those corresponding to DML ¼ 1) are circularly polarized in opposite senses. These lines are called the -lines. The central line (which is due to DML ¼ 0) is linearly polarized parallel to the applied field. It is called the p-line. The normal Zeeman effect is observed wherever spin is not present. It occurs even for transitions such as 1D ! 1P, in which the upper term is
7.20 THE ANOMALOUS ZEEMAN EFFECT
j
243
split into five states and the lower is split into three. In this case, the splittings are the same in the two terms, and the selection rules DML ¼ 0, 1 limit the transitions to three groups of coincident lines, as illustrated below. J L
S
m(spin)
Illustration 7.2. Analysing the splitting pattern for 1D ! 1P
Take the energy of an ‘unsplit’ P term to be zero. Then, in the presence of the magnetic field, the energies of ML ¼ 1, 0, þ1 are, by eqn 7.58, mB b; 0; mB b, respectively. Similarly, if the energy of an ‘unsplit’ D term is taken to be e, then the energies of the states ML ¼ 2, 1, 0, þ1, þ2 are e 2mB b; e mB b; e; e þ mB b; e þ 2mB b respectively. All DML ¼ þ1 transitions, for instance 1D(ML ¼ 2) ! 1P(ML ¼ 1), occur at an energy e mB b; all DML ¼ 0 transitions at e; and all DML ¼ 1 transitions at e þ mB b There are three groups, each one consisting of three coincident lines.
m (orbital) (a)
m (total)
7.20 The anomalous Zeeman effect
J L
S
m (spin)
m (orbital) (b)
m (total)
Fig. 7.22 (a) If the spin magnetic
moment of an electron bore the same relation to the spin as the orbital moment bears to the orbital angular momentum, the total magnetic moment would be collinear with the total angular momentum. (b) However, because the spin has an anomalous magnetic moment, the total moment is not collinear with the total angular momentum. The surviving component, after allowing for precession, is determined by the Lande´ g-factor.
The anomalous Zeeman effect, in which a more elaborate pattern of lines is observed, is in fact more common than the normal Zeeman effect. It is observed when the spin angular momentum is non-zero and stems from the unequal splitting of the energy levels in the two terms involved in the transition. That unequal splitting stems in turn from the anomalous magnetic moment of the electron (Section 7.3). If the g-value of an electron were 1 and not 2, then the total magnetic moment of the electron would be collinear with its total angular momentum (Fig. 7.22). But in fact, because of the anomaly, the two are not collinear. The spin and orbital angular momenta precess about their resultant (as a result of spin–orbit coupling), and as a result, the magnetic moment is swept around too. This motion has the effect of averaging to zero all except the component collinear with the direction of J, but the magnitude of this surviving magnetic moment depends on the values of L, S, and J because vectors of different lengths will lie at different angles to one another and give rise to different non-vanishing components of the angular momentum. The calculation of the surviving component of the magnetic moment runs as follows. The hamiltonian for the interaction of a magnetic field B with orbital and spin angular momenta is Hð1Þ ¼ morbital B mspin B ¼ ge ðL þ 2SÞ B
ð7:59Þ
where we have used 2 in place of ge. At this point, we look for a way of expressing the hamiltonian as proportional to J by writing Hð1Þ ¼ gJ ge J B
ð7:60Þ
where gJ is a constant. The two hamiltonians in eqns 7.59 and 7.60 are not equivalent in general, but for a first-order calculation we need only ensure that they have the same diagonal elements.
244
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
J
L
Consider Fig. 7.23. There are three precessional motions: S about J, L about J, and J about B. The effective magnetic moment can be found by projecting L on to J and then J on to B, and then doing the same for S. The precession averages to zero all the components perpendicular to this motion (this classical averaging is equivalent to ignoring all off-diagonal components in a quantum mechanical calculation). If k is a unit vector along J, it follows that the only surviving terms are L B ! ðL kÞðk BÞ ¼
S
S B ! ðS kÞðk BÞ ¼
k
ðL JÞðJ BÞ jJj2 ðS JÞðJ BÞ jJj2
Because J ¼ L þ S, it follows that Fig. 7.23 The vector diagram used
to calculate the Lande´ g-factor.
MJ
+3/2 2
D3/2
+1/2 –1/2 –3/2
2L J ¼ J2 þ L2 S2
If these quantities are now inserted into eqn 7.59 and the quantum mechanical expressions for magnitudes replace the classical values (so that J2 is replaced by J(J þ 1) h2, etc.), we find Hð1Þ ¼ ge ðL þ 2SÞ B
JðJ þ 1Þ þ SðS þ 1Þ LðL þ 1Þ JB ¼ ge 1 þ 2JðJ þ 1Þ This is the form we sought. It enables us to identify the Lande´ g-factor as gJ ðL; SÞ ¼ 1 þ
2
P1/2
+1/2 –1/2
Fig. 7.24 The anomalous Zeeman
effect. The splitting of energy levels with different g-values leads to a more complex pattern of lines than in the normal Zeeman effect.
2S J ¼ J2 þ S2 L2
JðJ þ 1Þ þ SðS þ 1Þ LðL þ 1Þ 2JðJ þ 1Þ
ð7:61Þ
When S ¼ 0, gJ ¼ 1 because then J must equal L. In this case, the magnetic moment is independent of L, and so all singlet terms are split to the same extent. This uniform splitting results in the normal Zeeman effect. When S 6¼ 0, the value of gJ depends on the values of L and S, and so different terms are split to different extents (Fig. 7.24). The selection rule DMJ ¼ 0, 1 continues to limit the transitions, but the lines no longer coincide and form three neat groups. Example 7.7 How to analyse the anomalous Zeeman effect
Account for the form of the Zeeman effect when a magnetic field is applied to the transition 2D3/2 ! 2P1/2. Method. Begin by calculating the Lande´ g-factor for each level, and then split
the states by an energy that is proportional to its g-value. Proceed to apply the selection rule DMJ ¼ 0, 1 to decide which transitions are allowed. Answer. For the level 2D3/2 we have L ¼ 2, S ¼ 12, and J ¼ 32. It follows that
g3/2(2, 12) ¼ 45. For the lower level, 2P1/2, we have g1/2(1, 12) ¼ 23. The splittings are
7.21 THE STARK EFFECT
j
245
therefore of magnitude 45mB b in the 2 D3=2 term and 23mB b in the 2 P1=2 term. The six allowed transitions are summarized in Fig. 7.24, where it is seen that they form three doublets. Self-test 7.7. Construct a diagram showing the form of the Zeeman effect
when a magnetic field is applied to a 3D2 ! 3P1 transition. J
L S
(a) L S
(b)
When the applied field is very strong, the coupling between L and S may be broken in favour of their direct coupling to the magnetic field.12 The individual angular momenta, and therefore their magnetic moments, now precess independently about the field direction (Fig. 7.25). As the electromagnetic field couples to the spatial distribution of the electrons (recall the form of the transition dipole moment), not to the magnetic moment due to the spin, the presence of the spin now makes no difference to the energies of the transitions. As a result, the anomalous Zeeman effect gives way to the normal Zeeman effect. This switch from the anomalous effect to the normal effect is called the Paschen–Back effect.
7.21 The Stark effect
Fig. 7.25 As the strength of the
applied field is increased, the precession of angular momenta about their resultant (as in (a)) gives way to precession about the magnetic field (as in (b)).
+ (b)
(a) Fig. 7.26 The origin of the first-order Stark effect. The two mixed states (a) and (b) give rise to two electron distributions that differ in energy.
The hamiltonian for the interaction with an electric field of strength e in the z-direction is Hð1Þ ¼ mz e ¼ eze
ð7:62Þ
where mz is the z-component of the electric dipole moment operator, mz ¼ ez. This operator has matrix elements between orbitals that differ in l by 1 but which have the same value of ml (recall Sections 5.16 and 7.2). The linear Stark effect is a modification of the spectrum that is proportional to the strength of the applied electric field. It arises when there is a degeneracy between the two wavefunctions that the perturbation mixes, as for the 2s and 2pz orbitals of hydrogen. The matrix element of the perturbation is, from Example 7.2, h2pz jHð1Þ j2si ¼ 3ea0 e
ð7:63Þ
and from Fig. 6.2 (or more formally from eqn 6.6) we know that the two degenerate orbitals mix and give rise to a splitting of magnitude 6ea0 e. The two functions that diagonalize the hamiltonian are N(2s 2pz), with N ¼ 1/21/2 (Fig. 7.26). It is easy to see that they correspond to a shift of charge density into and out of the direction of the field, and this difference in distribution accounts for their difference in energy. The splitting is .......................................................................................................
12. This feature is examined in Further information 15, where the full significance of the recoupling is seen to be the search for the representation that gave matrices with the smallest off-diagonal elements: the vector recoupling diagram is a pictorial representation of that effect.
246
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE Energy due to externally applied potential Coulomb potential energy
Total potential energy
very small: even for fields of 1.0 MV m1, the splitting corresponds to only 2.6 cm1. The linear Stark effect depends on the peculiar degeneracy characteristic of hydrogenic atoms, and is not observed for many-electron atoms where that degeneracy is absent. In these atoms, it is replaced by the quadratic Stark effect, which is even weaker. The origin of the effect is the same, but now the distortion of the charge distribution occurs only as a perturbation and the resulting shifts in energy are proportional to e2 . The field has to distort a nondegenerate and hence ‘tight’ system, and then interact with the dipole produced by that distortion. At very high field strengths the Ha line is seen to broaden and its intensity to decrease. These effects are traced to the tunnelling of the electron. In high fields the potential experienced by the electron has the form shown in Fig. 7.27. The tails of the atomic orbitals seep through the region of high potential and penetrate into the external region, where the potential can strip the electron away from the atom. This ionization results in fewer atoms being able to participate in emission, and so the intensity is decreased. Moreover, as the upper state has a shorter lifetime, its energy is less precise and the transition becomes diffuse.
Fig. 7.27 When the applied field is
very strong, its contribution to the total potential energy is such as to provide a tunnelling escape route for the originally bound state of an electron.
PROBLEMS 7.1 Calculate the wavenumbers of the transitions of He þ for the analogue of the Balmer series of hydrogen. Hint. Use eqn 7.2 with the Rydberg constant modified to account for the mass and charge differences. 7.2 Determine the longest possible wavelengths (the smallest wavenumbers) and the shortest possible wavelengths (the series limits) for lines in the (a) Lyman, (b) Balmer, (c) Paschen, and (d) Brackett series of the spectrum of atomic hydrogen. 7.3 Predict the form of the spectrum of the muonic atom formed from an electron in association with a m-meson (mm ¼ 207me, charge þ e).
7.6 Demonstrate that for one-electron atoms the selection rules are Dl ¼ 1, Dml ¼ 0, 1, and Dn unlimited. Hint. Evaluate the electric-dipole transition moment hn0 l0 m0l jjnlml i with mx ¼ er sin y cos f, my ¼ er sin y sin f, and mz ¼ er cos y. The easiest way of evaluating the angular integrals is to recognize that the components just listed are proportional to Ylml with l ¼ 1, and to analyse the resulting integral group theoretically. 7.7 Confirm that in hydrogenic atoms, the spin–orbit coupling constant depends on n and l as in eqn 7.24.
7.4 Which of the following transitions are electric-dipole allowed: (a) 1s ! 2s, (b) 1s ! 2p, (c) 2p ! 3d, (d) 3s ! 5d, (e) 3s ! 5p?
7.8 Calculate the spin-orbit coupling constant for a 2p-electron in a Slater-type atomic orbital, and evaluate it for the neutral atoms of Period 2 of the periodic table (from boron to fluorine).
7.5 The spectrum of a one-electron ion of an element showed that its ns-orbitals were at 0, 2 057 972 cm1, 2 439 156 cm1, and 2 572 563 cm1 for n ¼ 1, 2, 3, 4, respectively. Identify the species and predict the ionization energy of the ion.
7.9 Deduce the Lande´ interval rule, which states that for a given l and s, the energy difference between two levels differing in j by unity is proportional to j. Hint. Evaluate Eso in eqn 7.28 for j and j 1; use the second line in the equation (in terms of znl).
PROBLEMS
7.10 The ground-state configuration of an iron atom is 3d64s2, and the 5D term has five levels (J ¼ 4, 3, . . . , 0) at relative wavenumbers 0, 415.9, 704.0, 888.1, and 978.1 cm1. Investigate how well the Lande´ interval rule (Problem 7.9) is obeyed. Deduce a value of z3d. 7.11 (a) Calculate the energy difference between the levels with the greatest and smallest values of j for given l and s. Each term of a level is (2j þ 1)-fold degenerate. (b) Demonstrate that the barycentre (mean energy) of a term is the same as the energy in the absence of spin–orbit coupling. Hint. Weight each level with 2j þ 1 and sum the energies given in eqn 7.28 from j ¼ j l s j to j ¼ l þ s. Use the relations n X
s ¼ 12 nðn þ 1Þ
s¼0
n X
n X
s2 ¼ 16 nðn þ 1Þð2n þ 1Þ
s¼0
s3 ¼ 14 n2 ðn þ 1Þ2
s¼0
7.12 Identify the terms that may arise from the ground configurations of the atoms of elements of Period 2 and suggest the order of their energies. Hint. Construct the term symbols as explained in Section 7.6 and use Hund’s rules to arrive at their relative orders. Recall the hole–particle rule explained in Example 7.5. 7.13 Find the first-order corrections to the energies of the hydrogen atom that result from the relativistic mass increase of the electron. Hint. The energy is related to the momentum by E ¼ (p2c2 þ m2c4)1/2 þ V. When p2c2 << m2c4, E mc2 þ p2/2m þ V p4/8m3c2, where the reduced mass m has replaced m. Ignore the rest energy mc2, which simply fixes the zero. The term p4/8m3c2 is a perturbation; hence calculate hnlml j H(1) j nlmli ¼ (1/2mc2)hnlml j (p2/2m)2 j nlmli ¼ (1/2mc2)hnlml j (Enlml V)2 j nlmli. We know Enlml ; therefore calculate the matrix elements of V ¼ e2/4pe0r and V2. 7.14 Write the hamiltonian for the lithium atom (Z ¼ 3) and confirm that when electron–electron repulsions are neglected the wavefunction can be written as a product c(1)c(2)c(3) of hydrogenic orbitals and the energy is a sum of the corresponding energies. 7.15 Write the explicit expression for the Slater determinant corresponding to the 1s22s1 ground state of atomic lithium. Demonstrate the antisymmetry of the wavefunction upon interchange of the labels of any two electrons. 7.16 The Slater atomic orbitals are normalized but not mutually orthogonal. In the Schmidt orthogonalization procedure one orbital c is made orthogonal to R another orbital c 0 by forming c00 ¼ c cc 0 , with c ¼ c c 0 dt.
j
247
Confirm that c00 and c 0 are orthogonal and construct a 2s-orbital that is orthogonal to a 1s-orbital from an STO basis. 7.17 Take a trial function for the helium atom as c ¼ c(1)c(2), with c(1) ¼ (z3/p)1/2e zrl and c(2) ¼ (z3/p)1/2e zr2, z being a parameter, and find the best ground-state energy for a function of this form, and the corresponding value of z. Calculate the first and second ionization energies. Hint. Use the variation theorem. All the integrals are standard; the electron repulsion term is calculated in Example 7.3. Interpret Z in terms of a shielding constant. The experimental ionization energies are 24.58 eV and 54.40 eV. 7.18 On the basis of the same kind of calculation as in Problem 7.17, but for general Z, account for the first ionization energies of the ions Liþ, Be2þ, B3þ, and C4þ. The experimental values are 73.5, 153, 258, and 389 eV, respectively. 7.19 Consider a one-dimensional square well containing two electrons. One electron has n ¼ 1 and the other has n ¼ 2. Plot a two-dimensional contour diagram of the probability distribution of the electrons when their spins are (a) parallel, (b) antiparallel. Devise a measure of the radius of the Fermi hole. Hint. Recall the discussion in Section 7.11. When the spins are parallel (for example, aa) the antisymmetric combination c1(1)c2(2) c2(1)c1(2) must be used, and when the spins are antiparallel, the symmetric combination must be used. In each case plot c2 against axes labelled x1 and x2. Computer graphics may be used to obtain striking diagrams, but a sketch is sufficient. 7.20 The first few S terms of helium lie at the following wavenumbers: 1s2 1S: 0; 1s12s1 1S: 166 272 cm1: 1s12s1 3S: 159 850 cm1; 1s13s1 1S: 184 859 cm1; 1s13s1 3S: 183 231 cm1. What are the values of K in the 1s12s1 and 1s13s1 configurations? 7.21 What levels may arise from the following terms: S, 2P, 3P, 3D, 2D, 1D, 4D? Arrange in order of increasing energy the terms that may arise from the following configurations: 1s12p1, 2p13p1, 3p13d1. What terms may arise from (a) a d2 configuration, (b) an f 2 configuration? 1
7.22 An excited state of atomic calcium has the electron configuration 1s22s22p63s23p63d14f 1. (a) Derive all the term symbols (with the appropriate specifications of S, L, and J) for the electron configuration. (b) Which term symbol corresponds to the lowest energy of this electron configuration? (c) Consider a 3F2 level of calcium derived from a different electron configuration than that shown above. Which of the term symbols determined in part (a) can participate in spectroscopic transitions to this 3F2 level?
248
j
7 ATOMIC SPECTRA AND ATOMIC STRUCTURE
7.23 Write down the Slater determinant for the ground term of the beryllium atom, and find an expression for its energy in terms of Coulomb and exchange integrals. Find expressions for the energy in terms of the Hartree–Fock expression, eqn 7.53. Hint. Use eqn 7.53 for the configuration 1s22s2; evaluate the expectation value hc j H j ci.
7.25 Calculate the Lande´ g-factor for (a) a term in which J has its maximum value for a given L and S; (b) a term in which J has its minimum value.
7.24 Calculate the magnetic field required to produce a splitting of 1 cm1 between the states of a 1P1 level.
7.27 Calculate the form of the spectrum for the Zeeman effect on a 3P ! 3S transition.
7.26 Transitions are observed and ascribed to 1F ! 1D. How many lines will be observed in a magnetic field of 4.0 T?
8
An introduction to molecular structure
The Born–Oppenheimer approximation 8.1 The formulation of the approximation 8.2 An application: the hydrogen molecule–ion Molecular orbital theory 8.3 Linear combinations of atomic orbitals 8.4 The hydrogen molecule 8.5 Configuration interaction 8.6 Diatomic molecules 8.7 Heteronuclear diatomic molecules
Now we come to the heart of chemistry. If we can understand the forces that hold atoms together in molecules, we may also be able to understand why, under certain conditions, initial arrangements of atoms change into new ones in the course of the events we call ‘chemical reactions’. The aim of this chapter is to introduce some of the features of valence theory, the theory of the formation of chemical bonds. The description of bonding has been greatly enriched by numerical techniques, and the following chapter describes these more quantitative aspects of the subject. There are two principal models of molecular structure: molecular orbital theory and valence bond theory. Both models contribute concepts to the everyday language of chemistry and so it is worthwhile to examine them both. However, molecular orbital theory has undergone much more development than valence bond theory, and we shall concentrate on it.
Molecular orbital theory of polyatomic molecules 8.8 Symmetry-adapted linear combinations 8.9 Conjugated p-systems 8.10 Ligand field theory 8.11 Further aspects of ligand field theory The band theory of solids 8.12 The tight-binding approximation 8.13 The Kronig–Penney model 8.14 Brillouin zones
The Born–Oppenheimer approximation It is an unfortunate fact that, having arrived in sight of the promised land, we are forced to make an approximation at the outset. Even the simplest molecule, Hþ ¨ dinger 2 , consists of three particles, and its Schro equation cannot be solved analytically. To overcome this difficulty, we adopt the Born–Oppenheimer approximation, which takes note of the great difference in masses of electrons and nuclei. Because of this difference, the electrons can respond almost instantaneously to displacement of the nuclei. Therefore, instead of trying to solve the Schro¨dinger equation for all the particles simultaneously, we regard the nuclei as fixed in position and solve the Schro¨dinger equation for the electrons in the static electric potential arising from the nuclei in that particular arrangement. Different arrangements of nuclei may then be adopted and the calculation repeated. The set of solutions so obtained allows us to construct the molecular potential energy curve of a diatomic molecule (Fig. 8.1), and in general a potential energy surface of a polyatomic species, and to identify the equilibrium conformation of the molecule with the lowest point on this curve (or surface). The Born–Oppenheimer approximation is very reliable for ground electronic states, but it is less reliable for excited states.
250
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
8.1 The formulation of the approximation
Equilibrium bond length Potential energy
Internuclear distance
The simplest approach to the formulation of the Born–Oppenheimer approximation is to consider a one-dimensional analogue of the hydrogen molecule–ion, in which all motion is confined to the z-axis (Fig. 8.2). The full hamiltonian, H, for the problem is H¼
X 2 q2 h h2 q2 þ Vðz, Z1 , Z2 Þ 2 2me qz 2mj qZ2j j¼1;2
ð8:1Þ
where z is the location of the electron and Zj, with j ¼ 1,2, the locations of the two nuclei. More simply: H ¼ Te þ TN þ V Fig. 8.1 A typical molecular
potential energy curve for a diatomic species.
for the electron kinetic energy, the nuclear kinetic energy, and the potential energy of the system, respectively. The Schro¨dinger equation is HCðz, Z1 , Z2 Þ ¼ ECðz , Z1 , Z2 Þ
ð8:2Þ
We attempt a solution of the form Cðz, Z1 , Z2 Þ ¼ cðz; Z1 , Z2 ÞwðZ1 , Z2 Þ e
1 X1
x
2
X2
Fig. 8.2 The coordinates used in the discussion of the Born– Oppenheimer approximation.
ð8:3Þ
where c is the electronic wavefunction and w (chi) is the nuclear wavefunction. The notation c(z;Z1,Z2) means that the wavefunction for the electron is a function of its position z and depends parametrically on the coordinates of the two nuclei in the sense that we get a different wavefunction c(z) for each arrangement of the nuclei. When this trial solution is substituted into eqn 8.2 we obtain Hcw ¼ wTe c þ cTN w þ Vcw þ W ¼ Ecw where X h 2 qc qw q2 c W¼ 2 þ w qZj qZj qZ2j 2mj j¼1;2
ð8:4Þ
!
This latter quantity is non-zero because c depends on the nuclear coordinates, so qc/qZj is non-zero. However, because the nuclear masses occur in the denominator, we suppose that W is small and can be neglected and instead of eqn 8.4 try to solve1 wTe c þ cTN w þ Vcw ¼ Ecw or, collecting terms and rearranging slightly, cTN w þ ðTe c þ VcÞw ¼ Ecw
ð8:5Þ
As a first step at solving eqn 8.5 we write Te c þ Vc ¼ Ee ðZ1 , Z2 Þc
ð8:6Þ
.......................................................................................................
1. W is responsible for so-called ‘non-adiabatic effects’, which can be very important when interactions between electronic states are significant. For further details, see the Further reading section.
8.2 AN APPLICATION: THE HYDROGEN MOLECULE–ION
j
251
for fixed values of the nuclear coordinates. This equation is the Schro¨dinger equation for the electron in a potential V that depends on the fixed locations of the two nuclei. The solution is the electronic wavefunction c, and the eigenvalue Ee(Z1,Z2) is the electronic contribution to the total energy of the molecule plus the potential energy of internuclear repulsion at the preselected nuclear locations. It is this function that when plotted against the nuclear position gives the molecular potential energy curve. Finally, on substituting eqn 8.6 into eqn 8.5, we find cTN w þ Ee cw ¼ Ecw and on cancelling c obtain TN w þ Ee w ¼ Ew
ð8:7Þ
This equation is the Schro¨dinger equation for the wavefunction w of the nuclei when the nuclear potential energy, now represented by Ee, has the form of the molecular potential energy curve. Its eigenvalue E is the total energy of the molecule within the Born–Oppenheimer approximation. From now on (in this chapter) we shall concentrate on eqn 8.6, but write it more simply and generally, and with the normal symbols for the potential energy and total energy, as Hc ¼ Ec
H¼
h 2 2 r þV 2me
ð8:8Þ
where V is the potential energy of the electron in the field of the stationary nuclei plus the nuclear interaction contribution and E is the total electronic and nucleus–nucleus repulsion energy for a stationary nuclear conformation. rA A
e
rB
8.2 An application: the hydrogen molecule–ion
B
R
Fig. 8.3 The coordinates
used to specify the hamiltonian for the hydrogen molecule–ion.
Even within the Born–Oppenheimer approximation there is only one molecular species for which the Schro¨dinger equation can be solved exactly: the hydrogen molecule–ion, Hþ 2 . The hamiltonian for this species is H¼
= constant = constant
Fig. 8.4 The elliptical coordinates m, n, and f used for the separation of variables in the exact treatment (within the Born–Oppenheimer approximation) of the hydrogen molecule–ion.
2 2 h e2 e2 e2 r þ 2me 4pe0 rA 4pe0 rB 4pe0 R
ð8:9Þ
with the distances defined in Fig. 8.3. The final term represents the repulsive interaction between the two nuclei, and within the Born–Oppenheimer approximation is a constant for a given relative location of the nuclei. As Hþ 2 has only one electron, it has a status in valence theory analogous to the hydrogen atom in the theory of atomic structure. Just as the Schro¨dinger equation for the hydrogen atom is separable and solvable when expressed in spherical polar coordinates, so the equation for Hþ 2 is separable and solvable when expressed in ‘ellipsoidal coordinates’ (m,n,f), where m¼
rA þ rB R
n¼
rA rB R
ð8:10Þ
and f is the azimuthal angle around the internuclear axis (Fig. 8.4). In these coordinates, the two nuclei lie at the foci of ellipses of constant m. The
252
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
1.4 1.2
6 2 5 4 1
Energy/hcR H
1.0 0.9 0.8
3
0.6 0.4 0.2
2
0
1
–0.2 0
4
8 R/a0
12
Fig. 8.5 The molecular potential energy curves for the hydrogen molecule–ion.
+ (a)
–
+
(b) Fig. 8.6 Contour diagrams of the (a) bonding and (b) antibonding orbitals (1s and 2s, respectively) of the hydrogen molecule–ion in the LCAO approximation.
resulting solutions are called molecular orbitals and resemble atomic orbitals but spread over both nuclei. The ‘exact’ molecular orbitals of Hþ 2 are mathematically much more complicated than the atomic orbitals of hydrogen, and as we shall shortly make yet another approximation, there is little point in giving their detailed form.2 However, some of their features are very important and will occur in other contexts. The molecular potential energy curves vary with internuclear distance, R, as shown in Fig. 8.5. The two lowest curves are of the greatest interest, and we concentrate on them. The steep rise in energy as R ! 0 is largely due to the increase in the nucleus–nucleus potential energy as the two nuclei are brought close together. At large distances, as R ! 1, the curves tend towards the values typical of a hydrogen atom with the second proton a long way away. The lowest curve passes through a minimum close to R ¼ 2a0, and its energy then lies about 0.20hcRH (2.7 eV) below the energy of a separated hydrogen atom and proton. This result suggests that Hþ 2 is a stable species (in the sense of having a lower energy than its dissociation products, but not in a chemical sense of being non-reactive), and that its bond length will be close to 2a0 (106 pm). The species is known spectroscopically: its minimum lies at 2.648 eV and its bond length is 106 pm, in very good agreement. The origin of the lowering of energy can be discovered by examining the form of the wavefunctions, but we have to be circumspect. Figure 8.6 shows the two molecular orbitals of lowest energy as contour diagrams for various values of R. The striking difference between them is that the higher energy orbital (denoted 2s) has an internuclear node whereas the lower energy orbital (1s) does not. There is therefore a much greater probability of finding the electron in the internuclear region if it is described by the wavefunction 1s than if it is described by 2s.3 The conventional argument then runs that because the electron can interact with both nuclei if its wavefunction is 1s, then it is in a favourable electrostatic environment and will have a lower energy than that of a separated hydrogen atom and proton. It is on the basis of such a simplistic argument that chemical bond formation is commonly associated with the accumulation of electron density in an internuclear region. The actual interpretation of the wavefunctions is, however, a much more delicate problem. The total energy of a molecule has contributions from several sources, including the kinetic energy of the electron. What appears to happen on bond formation (in Hþ 2 at least) is that, as R is reduced from a large value, the lowest energy wavefunction shrinks on to the nuclei slightly as well as accumulating in the internuclear region. The transfer of electron density into the internuclear region is disadvantageous, because it is removed from close to the nuclei. However, the shrinkage of the orbitals overcomes this disadvantage, for although a slight increase in kinetic energy accompanies the shrinking (because the wavefunction becomes more sharply .......................................................................................................
2. A reference to their form is provided in the Further reading section. 3. The label s signifies the cylindrical symmetry of the orbital about the internuclear axis. A s orbital has zero units of electronic orbital angular momentum about that axis, a fact used in Section 8.4.
8.3 LINEAR COMBINATIONS OF ATOMIC ORBITALS
j
253
curved), a significant reduction in potential energy overcomes all these unwanted effects, and the net outcome is a lowering of energy. The formation of 2s, on the other hand, results in a small expansion of the electron distribution around the nuclei, and that has a net energy-raising effect. In other words, it is not the shift of electron density into the internuclear region that lowers the energy of the molecule but the freedom that this redistribution gives for the wavefunction to shrink in the vicinity of the two nuclei. In what follows, we shall anticipate the formation of a bond—as signalled by a lowering of the energy of the molecule—whenever there is an enhanced probability density in the internuclear region, but accept that this might be no more than a correlation rather than a direct effect on the energy of the molecule. A detailed analysis has been performed only for H2þ , and the argument might be quite different in other molecules.4
Molecular orbital theory A difficulty will already have become apparent: the solution of the Schro¨dinger equation for H2þ is so complicated (even after making the Born– Oppenheimer approximation) that there can be little hope that exact solutions will be found for more complicated molecules. Therefore, we must resort to another approximation, but use the exact solutions for H2þ as a guide. Another reason why making a further approximation is quite sensible is that we already have available quite good atomic orbitals for many-electron atoms, and it seems appropriate to try to use them as a starting point for the description of many-electron molecules built from those atoms.
8.3 Linear combinations of atomic orbitals Inspection of the form of the wavefunctions for H2þ shown in Fig. 8.6 suggests that they can be simulated by forming linear combinations of hydrogen atomic orbitals: cþ fa þ fb
c fa fb
ð8:11Þ
where fa is a H1s-orbital on nucleus A and fb its analogue on nucleus B.5 In the first case, the accumulation of electron density in the internuclear region is simulated by the constructive interference that takes place between the two waves centred on neighbouring atoms. The nodal plane in the true wavefunction is recreated by the destructive interference between waves superimposed with opposite signs. The partial justification for simulating molecular orbitals as an LCAO, a linear combination of atomic orbitals, can be appreciated by examining .......................................................................................................
4. See M.J. Feinberg, K. Ruedenberg, and E.L. Mehler, The origin of binding and antibinding in the hydrogen moleculeion. Adv. Quantum Chem., 27, 5 (1970). 5. In this chapter, we use f to denote an atomic orbital and c to denote a molecular orbital.
254
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
the hamiltonian for the problem given in eqn 8.9. When the electron is close to nucleus A, rA rB, and the hamiltonian is approximately H¼
2 2 h e2 e2 r þ 2me 4pe0 rA 4pe0 R
ð8:12Þ
Apart from the final, constant term, this hamiltonian is the same as that for a hydrogen atom. Therefore, close to nucleus A, the wavefunction of the electron will resemble a hydrogen atomic orbital. The same is true close to B, and this form of the solution is captured by the two linear combinations constructed above. The same conclusions can be reached in a more formal way, one that is more readily extended to other species, by writing the molecular orbitals as the following LCAO: X cr fr ð8:13Þ c¼ r
The atomic orbitals used in this expansion constitute the basis set for the calculation. In principle, we should use an infinite basis set for a precise recreation of the molecular orbital, but in practice only a finite basis set is used. Throughout this chapter we shall assume that the members of the basis set are real and that each one is normalized to 1. The optimum values of the coefficients are found by applying the variation principle, which means (Section 6.10) that we have to solve the secular equations X cr fHrs ESrs g ¼ 0 ð8:14Þ r
where Hrs is a matrix element of the hamiltonian and Srs is an overlap matrix element. These secular equations have non-trivial solutions only if the secular determinant vanishes. We write this condition as jH ESj ¼ 0
ð8:15Þ
where H is the matrix of elements Hrs and S is the corresponding matrix of elements Srs. To make progress with finding the roots of the determinant j H ES j we need to evaluate the relevant matrix elements. For a basis set of two atomic orbitals, one on atom A and the other on an identical atom B, the secular determinant is 2 2 and we can write SAA ¼ SBB ¼ 1 SAB ¼ SBA ¼ S where S is the overlap integral, and HAA ¼ HBB ¼ a
HAB ¼ HBA ¼ b
where a is the molecular Coulomb integral and b is the resonance integral. The secular determinant is a E b ES b ES a E ¼ 0 and its roots are E ¼
ab 1S
ð8:16Þ
8.3 LINEAR COMBINATIONS OF ATOMIC ORBITALS
j
255
The corresponding values for the real coefficients of the normalized wavefunctions are cA ¼ cB cA ¼ cB
cA ¼ cA ¼
1 f2ð1 þ SÞg 1
1=2
f2ð1 SÞg1=2
for Eþ ¼
aþb 1þS
ab for E ¼ 1S
ð8:17Þ
We establish the detailed form of the Coulomb and resonance integrals as follows. First, we insert the explicit form of the hamiltonian into their definitions: 1 e2 e2 ð8:18Þ A A þ a ¼ hAjHjAi ¼ E1s rB 4pe0 4pe0 R where we have abbreviated j fAi by j Ai and will later abbreviate j fBi by j Bi. The combination e2/4pe0 will occur many times in the following and henceforth we denote it j0: j0 ¼
e2 4pe0
ð8:19Þ
in which case we can write 1 j0 a ¼ E1s j0 A A þ rB R j' A
B
Fig. 8.7 The interpretation of the integral j 0 as the total Coulombic potential energy arising from a charge distribution on A with nucleus B.
ð8:20Þ
The first term in this expression (E1s) is obtained because fA is an eigenfunction of the atomic hamiltonian. The second term corresponds to the total Coulombic energy of interaction between an electron density f2A and the second nucleus B (Fig. 8.7). We shall call this contribution j 0 : Z 2 fA j0 ¼ j0 dt ð8:21Þ rB This integral is positive. It follows that the total Coulomb integral is a ¼ E1s j0 þ
j0 R
ð8:22Þ
Example 8.1 The evaluation of overlap and Coulomb integrals
Evaluate (a) the overlap integral S and (b) the integral j 0 for an electron in the molecular orbital cþ composed of two hydrogenic 1s-orbitals. Method. To evaluate integrals of this kind, it is natural to use ellipsoidal
coordinates, eqn 8.10, for which the volume element is dt ¼ 18R3 ðm2 n2 Þ dmdndf with 1 m 1, 1 n 1, and 0 f 2p. The atomic wavefunctions we use are f(r) ¼ (Z3/pa30)1/2 eZr=a0 , where Z is the atomic number and a0 is the Bohr radius; each orbital is centred on its nucleus. All the integrations are straightforward in ellipsoidal coordinates.
j
256
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
1
Answer. (a) The product of the two wavefunctions is
0.8
fA ðrA ÞfB ðrB Þ ¼
0.6 S 0.4 0.2 0 0
2
R/a0
6
4
Fig. 8.8 The variation of the overlap integral of two H1s-orbitals with internuclear distance in the hydrogen molecule–ion.
1
a0j' /Zj0
0.8 0.6
3 Z3 ZðrA þrB Þ=a0 Z e eZmR=a0 ¼ pa30 pa30
Therefore, the overlap integral is Z Z Z Z3 R3 2p 1 1 2 S ¼ hAjBi ¼ ðm n2 ÞeZmR=a0 dmdndf 3 8pa0 0 1 1 Z 1Z 1 Z3 R3 ¼ 2p ðm2 n2 ÞeZmR=a0 dmdn (after integration over f) 8pa30 1 1 Z 1 Z3 R3 2 ZmR=a0 2 e 2p 2m dm (after integration over n) ¼ 3 8pa3 1 ( 0 ) ZR 1 ZR 2 ZR=a0 ¼ 1þ þ e a0 3 a0 (b) The contribution j 0 is similarly, using m2 n2 ¼ (m þ n)(m n):
Z Z Z j0 Z3 2p 1 1 18 R3 m2 n2 eZðmþnÞR=2a0 j0 ¼ dmdndf 1 pa30 0 1 1 2 Rðm nÞ 3 Z 1 Z 1 1 Z R2 ðm þ nÞeZðmþnÞR=2a0 dmdn ¼ j0 2 a0 1 1 1 ZR 2ZR=a0 e ¼ j0 1 1 þ R a0 These two functions are plotted in Figs 8.8
0.4
and 8.9
.
0
Comment. Both S and j decrease as Z increases because the higher nuclear
0.2 0 0
2
R/a0
6
4
charge shrinks the orbitals down on to their respective nuclei. A more detailed account of the calculation of molecular integrals is given in S.P. McGlynn, L.G. Vanquickenborne, M. Kinoshita, and D.G. Carroll, Introduction to applied quantum chemistry, Holt, Rinehart, and Winston, New York (1972). Self-test 8.1. Evaluate the overlap integral between two Slater 2s-orbitals on
Fig. 8.9 The variation of the integral
different atoms.
j 0 with internuclear distance in the hydrogen molecule–ion.
For the resonance integral b, we use the fact that fB is an eigenfunction of the hamiltonian for hydrogen atom B with eigenvalue E1s, and write 1 j0 b ¼ hAjHjBi ¼ E1s hAjBi j0 A B þ hAjBi rA R ð8:23aÞ j0 S k0 ¼ E1s þ R
k' A
AB
B
where 0
k ¼ j0 Fig. 8.10 The interpretation of the
integral k 0 as the interaction of an overlap charge distribution with one of the nuclei.
Z
fA fB dt rA
The analytical expression for k 0 for two H1s-orbitals is j0 R R=a0 0 e k ¼ 1þ a0 a0
ð8:23bÞ
ð8:24Þ
8.3 LINEAR COMBINATIONS OF ATOMIC ORBITALS
Energy
E– A E1s E+
– B E1s
+
energy level diagram of the hydrogen molecule–ion in the LCAO approximation. Note that the 2s-orbital is slightly more antibonding than the 1s-orbital is bonding. 5 4 3 2
a0E/j0
0.1
1 0
–0.1
0
0
5 R /a0
10
Fig. 8.12 The calculated molecular
potential energy curves of the two lowest energy molecular orbitals of the hydrogen molecule–ion within the LCAO approximation. Note the change in scale between the bonding and antibonding curves.
+ (a)
– (b)
257
The integral k 0 , which in this case is positive, has no classical analogue. However, an indication of its significance is that we can think of it as representing the interaction of the overlap charge density, efAfB, with nucleus A (Fig. 8.10). By symmetry, the interaction with nucleus B has the same value. It follows that the energies of the two LCAO-MOs are j0 j0 þ k0 R 1þS j0 j0 k0 E ¼ E1s þ R 1S Eþ ¼ E1s þ
Fig. 8.11 The molecular orbital
0.2
j
+ +
Fig. 8.13 The parity classification of
orbitals in a homonuclear diatomic molecule: (a) g, (b) u.
ð8:25Þ
The integrals j 0 and k 0 are both positive, with j 0 > k 0 ; the lower of the two energies is E þ (Fig. 8.11). The ladder of energy levels is called a molecular orbital energy level diagram. The lower-energy orbital, which has the form c þ ¼ fA þ fB, is called a bonding orbital. The higher-energy orbital, which has the form c ¼ fA fB, is called an antibonding orbital. Occupation of a bonding orbital lowers the energy of a molecule and helps to draw the two nuclei together; when an antibonding orbital is occupied, the energy of the molecule is raised and the two nuclei tend to be forced apart. One feature that should be noticed is that the diagram is not quite symmetrical: the antibonding orbital lies further above the energy of a hydrogen atom than the bonding orbital lies below it (as shown in Figs 8.5 and 8.11). This asymmetry is largely due to the repulsion between the two nuclei, which pushes both orbitals up in energy. In other words, an antibonding orbital is more antibonding than a bonding orbital is bonding. The analytical expressions for the energies are plotted in Fig. 8.12. As can be seen, the molecular potential energy curve has a minimum close to R ¼ 2.5a0 (130 pm) at a depth of 0.13hcRH (170 kJ mol 1). The experimental values are 2.0a0 (106 pm) and 0.195hcRH (255 kJ mol 1), and so the agreement is not spectacularly good. The principal source of error is that the basis is insufficiently flexible: we saw in the discussion of the exact solutions that a major contribution to the bonding comes from a shrinkage of the orbitals on to their respective nuclei, but this feature cannot be captured by the present model. One final detail of the molecular orbitals can usefully be introduced at this stage. The two molecular orbitals we have constructed can be classified according to their parity, their symmetry properties under inversion of the electron coordinates. As indicated in Fig. 8.13, under inversion the wavefunction c þ remains indistinguishable from itself, and hence it is classified as having gerade symmetry, denoted g, where gerade is the German word meaning ‘even’. In contrast, c changes sign under inversion, so it is classified as ungerade, the German word for ‘odd’, and denoted u. The full-dress versions of the orbital labels are therefore 1sg and 1su (note that when u and g are added as labels, each set is labelled separately). We already know that the symmetry classification is important for the discussion of selection rules (Section 7.2); we shall see that the same classification also helps us to understand the electronic structures of molecules.
258
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
8.4 The hydrogen molecule We model the electronic structure of the hydrogen molecule, H2, by the addition of a second electron to the 1sg orbital, to give the configuration 1s2g . The orbital description is therefore cþ(1)cþ(2), where the 1 and 2 in parentheses are short for r1 and r2, respectively, the locations of the two electrons. Writing the true wavefuntion c(1,2) as a product is an approximation that is valid only if electron–electron interactions are ignored or replaced by some kind of average one-electron potential energy (as in the central-field approximation, Section 7.12) so that the true hamiltonian 2 2 h h2 2 e2 e2 r1 r2 2me 2me 4pe0 rA1 4pe0 rB1 e2 e2 e2 e2 þ þ 4pe0 rA2 4pe0 rB2 4pe0 r12 4pe0 R
H¼
ð8:26aÞ
is replaced by an expression of the form H ¼ H1 þ H2 þ
e2 4pe0 R
ð8:26bÞ
where each Hi is expressed in terms of the coordinates of the electron i alone. The approximate spatial wavefunction cþ(1)cþ(2) is symmetric under particle interchange, so the spin component must be proportional to a(1)b(2) b(1)a(2) to guarantee that the overall wavefunction is antisymmetrical. Therefore, when the two electrons enter a single molecular orbital, they do so with paired spins ("#). Spin-pairing is thus seen not to be an end in itself, but the way that electrons must arrange themselves in order to pack into the lowest energy orbital. The ground-state configuration of H2 is classified as 1Sg in an echo of the term symbols used for atoms. The superscript 1 is the spin multiplicity of the state, which in this instance corresponds to S ¼ 0 because the two electrons are paired.6 The S (uppercase sigma) is the analogue of the letter S used to denote full spherical rotational symmetry and indicates that the total orbital angular momentum around the internuclear axis is zero because both electrons occupy s-orbitals, and so neither has orbital angular momentum about the axis. More formally, we denote the component of orbital angular momentum about the axis as l for each electron, and the total as L ¼ l1 þ l2. In ground-state H2, l1 ¼ l2 ¼ 0, so L ¼ 0, corresponding to a S term. The subscript g indicates that the overall parity of the state is g. To calculate it from the individual values for each electron we use gg¼g
gu¼u
uu¼g
which follow from the mathematical properties of the products of odd and even functions, and use the first of these results for this two-electron system in .......................................................................................................
6. As usual, we run into a paucity of letters. Be careful to distinguish S (upper case italic) for the overlap integral, S (upper case italic) for the total spin quantum number, and S (upper case roman) for a symmetry label of an atomic term.
8.5 CONFIGURATION INTERACTION
If f(x) is an even function, so that f(x) ¼ f(x), and g(y) is an odd function, so that g(y) ¼ g(y), then the product h(x,y) ¼ f(x)g(y) is an odd function: hðx, yÞ ¼ f ðxÞgðyÞ ¼ f ðxÞgðyÞ ¼ hðx, yÞ
j
259
which both electrons occupy g orbitals. Had one electron occupied a su orbital, then the term would have been of overall u parity. The full form of the H2 two-electron approximate wavefunction is cð1, 2Þ ¼ cþ ð1Þcþ ð2Þs ð1, 2Þ
ð8:27Þ
where the factor s is the spin contribution (12)1/2{a(1)b(2) a(2)b(1)}. The energy of the molecule is found by evaluating the expectation value of the hamiltonian in eqn 8.26a. The resulting expression is E ¼ 2E1s þ
j0 2j0 þ 2k0 j þ 2k þ m þ 4l þ R 1þS 2ð1 þ SÞ2
ð8:28Þ
where, in addition to the integrals already defined, we need: The repulsion of a charge density of electron 1 on A with the charge density of electron 2 on B: Z 1 j ¼ j0 fA ð1Þ2 f ð2Þ2 dt1 dt2 ¼ ðAAjBBÞ ð8:29aÞ r12 B The repulsion of the overlap charge density of electron 1 and the overlap charge density of electron 2: Z 1 k ¼ j0 fA ð1ÞfB ð1Þ f ð2ÞfB ð2Þdt1 dt2 ¼ ðABjABÞ ð8:29bÞ r12 A The repulsion of the charge density of electron 1 on A with the overlap charge density of electron 2: Z 1 f ð2ÞfB ð2Þdt1 dt2 ¼ ðAAjABÞ ð8:29cÞ l ¼ j0 fA ð1Þ2 r12 A
0.4
The repulsion of the charge density of electron 1 on A with the charge density of electron 2 also on A: Z 1 m ¼ j0 fA ð1Þ2 f ð2Þ2 dt1 dt2 ¼ ðAAjAAÞ ð8:29dÞ r12 A
(E – 2E1s)/ hcRH
0.2
0
1.4
–0.2 –0.27
Experimental
–0.33 –0.4 0
1
2 R /a0
3
Fig. 8.14 The calculated molecular
potential energy curve for the lowest energy orbital of a hydrogen molecule in the LCAO approximation.
The values for these integrals are given in Further information 10; the notation on the right will be used again in Chapter 9. The molecular potential energy curve calculated from this expression is shown in Fig. 8.14. It has a minimum at R ¼ 1.4a0 (74 pm), and the minimum lies at 0.27hcRH (350 kJ mol 1) below 2E1s, the energy of two separated hydrogen atoms. The experimental values are 1.40a0 (74.1 pm) and 0.33hcRH (430 kJ mol1), respectively, and although there is a fair measure of agreement, there is room for improvement. The kind of improvement that can be made includes the use of a more flexible basis set, the use of SCF procedures, and the incorporation of electron correlation (see Chapter 9).
8.5 Configuration interaction One procedure that is widely used and can be illustrated here is the method of configuration interaction (CI), first mentioned in connection with atoms in Section 7.17. If we continue to stick with the two-orbital basis set,
j
260
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
then there are two molecular orbitals for the two electrons of H2 to occupy. The following configurations are possible:
3
1s2g (E – 2E1s)/hcRH
1s2u
The corresponding wavefunctions, including spin, are
2
C1 ð1, 2; 1 Sg Þ ¼ cþ ð1Þcþ ð2Þs ð1, 2Þ
g
1
C2 ð1, 2; 1 Su Þ ¼ ð12Þ1=2 fcþ ð1Þc ð2Þ þ cþ ð2Þc ð1Þgs ð1, 2Þ
u
1
1
C3 ð1, 2; 1 Sg Þ ¼ c ð1Þc ð2Þs ð1, 2Þ
u
3
C4 ð1, 2; 3 Su Þ ¼ ð12Þ1=2 fcþ ð1Þc ð2Þ cþ ð2Þc ð1Þgsþ ð1, 2Þ
g
1
0
0
1s1g 1s1u
1
2 R /a0
3
4
Fig. 8.15 The variation of the
energies of four states of the hydrogen molecule with changing internuclear distance and the effect of configuration interaction which pushes the two pale curves apart.
Each state has been classified according to the procedure indicated in Section 8.4 and has been constructed to be antisymmetric with respect to electron interchange, as required by the Pauli principle. The MO description given above considered only C1(1,2). When the energies of all four terms are calculated, we obtain the molecular potential energy curves shown in Fig. 8.15. One important feature is that two of them, C1(1,2) and C3(1,2), converge on the same energy as R ! 1. Moreover, they are both 1Sg terms. We have already seen that states of the same symmetry never cross because the hamiltonian always has non-zero matrix elements between them.7 As a result configuration interaction occurs, and instead of crossing the two terms move apart as shown in Fig. 8.15. Configuration interaction lowers the energy of the lower term because their interaction in effect pushes the two states apart (as in Fig. 6.1), and hence leads to an improved description of the ground state and a lowering of its energy. With CI, the wavefunction of the lower state is Cð1, 2Þ ¼ c1 C1 ð1, 2Þ þ c3 C3 ð1, 2Þ ¼ Fð1, 2Þs ð1, 2Þ
ð8:30Þ
and the orbital structure of this function is Fð1, 2Þ ¼ c1 cþ ð1Þcþ ð2Þ þ c3 c ð1Þc ð2Þ ¼ 12 c1 ffA ð1Þ þ fB ð1ÞgffA ð2Þ þ fB ð2Þg þ 12 c3 ffA ð1Þ fB ð1ÞgffA ð2Þ fB ð2Þg ¼ 12 ðc1 þ c3 ÞffA ð1ÞfA ð2Þ þ fB ð2ÞfB ð1Þg þ 12ðc1 c3 ÞffA ð1ÞfB ð2Þ þ fB ð1ÞfA ð2Þg
ð8:31aÞ
It is revealing to compare this wavefunction with the form it has in the absence of CI (setting c1 ¼ 1 and c3 ¼ 0): Fð1, 2Þ ¼ 12 fA ð1ÞfA ð2Þ þ 12 fB ð2ÞfB ð1Þ þ 12 fA ð1ÞfB ð2Þ þ 12 fB ð1ÞfA ð2Þ
ð8:31bÞ
The key point is that the former wavefunction is more flexible because the coefficients c1 þ c3 and c1 c3 are variable; there is no such flexibility in .......................................................................................................
7. In diatomic molecules, potential energy curves for states of the same symmetry do not intersect but rather undergo avoided crossings. In polyatomic molecules, potential energy surfaces of electronic states of the same symmetry may intersect; one important and common type of intersection of electronic states is the so-called ‘conical intersection’.
8.6 DIATOMIC MOLECULES
j
261
the latter wavefunction. This relaxation of constraint is an improvement and is reflected in the lowering of energy of the lower of the two interacting states.
8.6 Diatomic molecules
(a)
(b)
(c)
(d)
Fig. 8.16 Examples of varieties
of molecular orbitals: (a) and (b) s-orbitals, (c) and (d) p-orbitals.
It is not a long step, at least at the present level of exposition, from H2 to the LCAO-MO description of other homonuclear diatomic molecules. The basic principle for the construction of molecular orbitals is to form linear combinations of atomic orbitals that have the same symmetry with respect to rotations about the internuclear axis. More formally, we build linear combinations of atomic orbitals that have the same symmetry species (that is, span the same irreducible representation) within the molecular point group. As we established in Section 5.16, only orbitals of the same symmetry species may have non-zero overlap (S 6¼ 0) and hence only they contribute to bonding. Thus, with the internuclear axis taken as the z-axis, s-, pz-, and dz2-orbitals all have symmetry species S in C1v and may contribute to s-orbitals. Similarly, px- and py-orbitals jointly span P in C1v, and hence may contribute to porbitals (Fig. 8.16) which have one unit of orbital angular momentum about the internuclear axis; dyz- and dzx-orbitals also span irreducible representations of symmetry species P, and they too may contribute to p-orbitals. It is rarely necessary to consider d-orbitals, but the same principles can be applied: we select atomic orbitals of symmetry species D (specifically dxy and dx2 y2 ), and form linear combinations of them. We have stressed that group theory provides techniques for selecting atomic orbitals that may contribute to bonding, but other types of arguments must be used to decide whether these orbitals do in fact contribute, and to what extent. There are essentially two criteria to keep in mind in this connection. First, to participate significantly in bond formation, atomic orbitals must be neither too diffuse nor too compact. In either case, there would be only weak constructive or destructive overlap between neighbouring atoms, and only feeble bonds would result. It follows that in Period 2, (1s,1s)-overlap can be largely neglected in comparison with (2s,2s)-overlap, for 1s-orbitals are too compact to have significant overlap with each other. Indeed, it is generally safe, for qualitative discussions at least, to consider only overlap between orbitals of the valence shell, for only these orbitals are neither too compact nor too diffuse to have significant overlap. Second, the energies of the orbitals should be similar. To see why this is so, consider the following secular determinant for the bond formed between two different atoms A and B: aA E b ES ð8:32Þ b ES aB E ¼ 0 The roots are found by solving the quadratic equation for the energy, and when j aA aB j b and S ¼ 0 they are Eþ ¼ aA
b2 aB aA
E ¼ aB þ
b2 aB aA
ð8:33Þ
j
262
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
A
B
(a)
/( A – B ) 2
B
A
2/( A – B ) (b) Fig. 8.17 The molecular orbital
energy levels stemming from atomic orbitals of (a) the same energy, (b) different energy.
These results (which are illustrated in Fig. 8.17) show that the molecular orbital energies are shifted from the atomic orbital energies (aA and aB) by only a small amount when aA and aB are very different. The implication is that in homonuclear diatomic molecules, the atomic orbitals of identical energy dominate the bonding. The strongest bonds will therefore have compositions such as (2s,2s) and (2p,2p), and there is no need (for qualitative discussions, at least) to consider (2s,1s) and (2p,1s) contributions. There is normally insufficient energy difference between 2s- and 2p-orbitals for it to be safe to ignore (2s,2p) contributions, although in elementary accounts that is often adopted as an initial approximation. With these rules in mind, it is quite easy to set up a plausible molecular orbital energy level diagram for the Period 2 homonuclear diatomic molecules. We consider only the valence orbitals (and, in due course, the electrons they contain). From the four atomic orbitals of S symmetry (the 2s- and 2pz-orbitals on each atom), we can form four linear combinations; these are the four s-orbitals marked on the diagram. To a first approximation we can think of the 2s-orbitals as forming bonding and antibonding combinations and the 2pz-orbitals as doing the same. However, it is better to think of all four combinations as formed from the four atomic orbitals, with increasing energy from the most bonding combination (1sg) to the most antibonding combination (2su). All four s-orbitals have mixed 2s- and 2pz-orbital character, with the lowest energy combination predominantly 2s-orbital in character and the highest energy combination predominantly 2pz. The four orbitals with P symmetry likewise form four combinations, but because they span the two-dimensional irreducible representation, they fall into two doubly degenerate sets, which we call 1pu and 1pg. It is hard to predict the order of energy levels, particularly the relative ordering of the s and p sets, but it is found experimentally and confirmed by more detailed calculations that the order shown on the left of Fig. 8.18 applies from Li2 to N2, whereas the order shown on the right applies to O2 and F2. To arrive at the electron configuration of the neutral molecule, we add the appropriate number of valence electrons to each set of energy levels. The procedure mirrors the building-up principle for atoms in that the electrons are added to the lowest energy available orbital subject to the requirement of the Pauli exclusion principle. If more than one orbital is available (as is the case when electrons occupy the p-orbitals), then electrons first occupy separate orbitals so as to minimize electron–electron repulsions; moreover, to benefit from spin correlation (Section 7.11), they do so with parallel spins. For instance, for N2 we need to accommodate 10 valence electrons, and the ground-state configuration is N2
1s2g 1s2u 1p4u 2s2g
1
Sg
For O2, there are 12 valence electrons to accommodate, and the expected configuration is O2
1s2g 1s2u 2s2g 1p4u 1p2g
3
Sg
8.6 DIATOMIC MOLECULES
j
263
Energy 2 u
1g
2g
Ne2
2 u
O2, F2
1g
N2
2p
2p
1u
B2, C 2
2g
Fig. 8.18 The molecular orbital
energy levels of diatomic molecules of the Period 2 elements. The labels indicate the highest occupied level of the specified species. Note the change in the order that appears between dinitrogen and dioxygen.
1u
1u
Be2
1u 2s
2s 1g
Li2
1g
Note that because only two electrons occupy the 1pg orbital, they will be in separate orbitals and have parallel spins. Hence the ground state is predicted to be a triplet (S ¼ 1). The possession of non-zero spin is consistent with the paramagnetic character of oxygen gas. The terms HOMO and LUMO are used to refer to the highest occupied and lowest unoccupied molecular orbitals, respectively, which in the case of O2 are the 1pg (HOMO) and 2su (LUMO) orbitals. The HOMO and LUMO are referred to jointly as the frontier orbitals: the frontier is the site of much of the reactive and spectroscopic activity of the species. The term symbols that have been attached to the configurations listed above have been worked out in the manner already sketched for H2. However, the symbol for O2 is quite instructive (and incomplete). To determine the value of L, the total orbital angular momentum around the internuclear axis, we add together all the individual ls. For s-orbitals (and each electron they contain), l ¼ 0. For p-orbitals, l ¼ 1 because each orbital corresponds to a different sense of rotation about the axis: l ¼ þ1 corresponds to the linear combination p þ ¼ px þ ipy and l ¼ 1 corresponds to p ¼ px ipy (in each case the unwritten normalization factor is 1/21/2). It follows that a p4 configuration necessarily contributes 0 to L, because it has equal numbers of electrons orbiting clockwise and counter-clockwise. A p2 configuration, as in O2, however, can contribute 0 or 2 because the two electrons can be in different orbitals (p1þ p1 ) or the same orbital (p2þ or p2 ), respectively. Hence, the configuration can give rise to a S ( j L j ¼ 0) and a D ( j L j ¼ 2) term, respectively. Because we expect the two electrons to occupy different orbitals (to minimize their mutual repulsion), it follows that we expect the ground term to be S, with D higher in energy. Moreover, because the electrons are in different p-orbitals, they can have either S ¼ 0 or S ¼ 1, so we
264
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
expect 1S and 3S terms, with the latter lower in energy. On the other hand, the D term must have paired spins because both electrons occupy the same p-orbital, and so it must have S ¼ 0, so giving a 1D term. The experimental molecular potential energy curves for O2 are illustrated in Fig. 8.19, and these terms can be identified. We remarked that the term symbol for O2 given above is incomplete. Terms designated S also require a label to distinguish their behaviour under reflection in a plane that contains the internuclear axis (see the C1v character table in Appendix 1). Each s-orbital has the character þ1 under this operation. A p-orbital, however, may have the character þ1 or 1 (Fig. 8.20). If the two electrons of interest occupy different p-orbitals, then one of them will be þ1 and the other will be 1, and overall the character of the configuration will be ( þ1) ( 1) ¼ 1. This symmetry is denoted by a right superscript, so the full term symbol for the ground state of O2 is 3Sg . The case of C2 is equally instructive. The straightforward application of the building-up principle suggests the ground-state configuration C2
1s2g 1s2u 1p4u
1
Sþ g
However, we have to be circumspect, because we are dealing with a manyelectron molecule, and the occupation of the lowest energy orbitals does not necessarily lead to the lowest energy. We need to allow for the possibility that excitation of an electron to a nearby orbital, as illustrated in Fig. 8.21, might lower the electron–electron repulsion and result in a lower overall energy despite the occupation of a higher energy orbital. The resulting configuration1s2g 1s2u 1p3u 2s1g would result in a 3Pu term, with the lowering of energy aided by the presence of spin correlation. Provided that the 1pu and (a) 2g
4 Σg+
1
2
1
O( P) + O( D)
Σu+
O(3P) + O(3P)
3
ñ
1πu
E/eV
0
3
Σu
3
–2
(a)
Σg+
1
∆g
1
3
–4
–6
100
Σ
3
ñ u
200 R/pm
(b)
2g
Πu 1πu
Fig. 8.21 When orbitals have similar
300 (b)
Fig. 8.19 The experimentally
determined molecular potential energy curves of some of the lower energy states of dioxygen.
Fig. 8.20 The origin of the
þ/ symmetry classification: (a) a p_-orbital, (b) a p þ -orbital.
energies, there may be a competition to determine whether (a) the lowest energy orbitals are occupied or (b) a higher energy orbital is occupied, with the advantage of the effects of spin correlation.
8.7 HETERONUCLEAR DIATOMIC MOLECULES
j
265
2sg orbitals are quite close in energy, there is no unambiguous way of predicting which is the lower state. Indeed, even the experimental situation was unclear for many years, but it has now been resolved in favour of a 1Sþ g ground state. This qualitative approach to the electronic structure of diatomic molecules is only a first stage in reaching an understanding. Modern quantitative theories of structure are based on detailed numerical calculations like those described in Chapter 9.
8.7 Heteronuclear diatomic molecules The qualitative effect of the presence of two different atoms in a molecule is for there to be a non-uniform distribution of electron density. Specifically, for molecular orbitals of the form c ¼ cA f A þ cB f B
ð8:34Þ 2
4 C2p
2
O2p 3 1 C2s 2 O2s 1 Fig. 8.22 A schematic depiction of
the molecular orbital energy levels of the carbon monoxide molecule.
2
it will no longer be true that j cA j ¼ j cB j . A useful rule of thumb is that if A is the more electronegative atom of the two, then the bonding combination will have j cA j 2 > j cB j 2, with cA and cB of the same sign, as it is a contribution to the lowering of energy for the electron to be found predominantly on A. On the other hand, for the antibonding combination, j cA j 2 < j cB j 2, with cA and cB of opposite sign, and an electron in this orbital will be found predominantly on B. Its occupation of an orbital on B is a contribution to the raising of the energy of this molecular orbital. A second feature of heteronuclear bonding is that because, except by accident, the energies of the orbitals of one atom do not coincide with those of the second atom, the extent to which the molecular orbitals are shifted in energy from the atomic orbitals is less than for homonuclear species. To borrow a term from classical physics, in homonuclear molecules the orbitals of the same designation ‘resonate’ with one another and hence couple strongly, whereas the resonance is imperfect in heteronuclear species and the coupling is weaker. These features suggest that a heteronuclear bonding system can be generated from the homonuclear system by reducing the shifts in energy represented by the molecular orbital energy levels, and moving bonding orbitals towards the lower energy contributing atomic orbitals to represent their greater contribution; antibonding orbitals are similarly shifted towards the higher energy orbitals. The resulting scheme (for CO) is illustrated in Fig. 8.22. Note also that, because a heteronuclear diatomic molecule lacks a centre of inversion, the parity designation (g or u) is no longer relevant. The electron configuration of CO can now be deduced by adding the 10 valence electrons to the lowest five orbitals: CO
1s2 2s2 1p4 3s2
1
Sþ
The HOMO is 3s, which is in fact a largely non-bonding orbital on the C atom, so 3s2 corresponds to a lone pair on C. The LUMO is 2p, which is a doubly degenerate pair of orbitals of largely C2p-orbital character. This combination of a HOMO that can provide two electrons and a LUMO that
266
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
can accept them is potent, and accounts for the widespread occurrence of metal carbonyl complexes such as Ni(CO)4 and for the ability of carbon monoxide to act as a poison. Once again, it must be stressed that arguments such as these are little more than a qualitative rule of thumb for rationalizing certain features of the electronic structures of diatomic molecules. For accurate energies and electron distributions, and to calculate reliable molecular properties from these wavefunctions, it is necessary to use the numerical techniques described in Chapter 9.
Molecular orbital theory of polyatomic molecules The molecular orbitals of polyatomic species are LCAOs just like those in eqn 8.13, which we repeat here: X c¼ cr fr ð8:35Þ r
The main difference is that now the sum extends over all the atomic orbitals of the atoms in the molecule. However, as for diatomic molecules, only atomic orbitals that have the appropriate symmetry make a contribution, because only they have net overlap with one another. When a molecule lacks any symmetry elements (other than the identity), there is no way of avoiding assembling each molecular orbital from the entire basis set. However, when the molecule has elements of symmetry, group theory can be particularly helpful in deciding which orbitals can contribute to each molecular orbital, and in classifying the resulting orbitals according to their symmetry species.
8.8 Symmetry-adapted linear combinations The concept behind the construction of a symmetry-adapted linear combination (SALC) is to identify two or more equivalent atoms in a molecule, such as the two H atoms in H2O, and to form linear combinations of the atomic orbitals they provide that belong to specific symmetry species. Then molecular orbitals are constructed by forming linear combinations of each SALC with an atomic orbital of the same symmetry species on the central atom (the O atom in H2O). We can be confident that only the SALC with a given symmetry species will have a net overlap with an atomic orbital of the same symmetry species. The effect of using SALCs instead of the raw basis is to factorize the secular determinant into block-diagonal form, because all elements Hrs and Srs are zero except between orbitals of the same symmetry species. The secular determinant is thereby factorized into a product of smaller determinants, and we need to find the roots of these determinants, which is in general a much simpler task. An example should make this clear. Consider the H2O molecule, which belongs to the point group C2v. If we use the six-member basis (H1sA, H1sB, O2s, O2px, O2py, O2pz), then we should expect a 6 6 determinant and
8.8 SYMMETRY-ADAPTED LINEAR COMBINATIONS
j
267
a sixth-order equation to solve for E. However, it should be clear from Fig. 8.23 that the two linear combinations a1
O2s, O2pz
fðA1 Þ ¼ H1sA þ H1sB
fðB2 Þ ¼ H1sA H1sB
can have net overlap with O2s and O2pz (for f(A1) ) and with O2py (for f(B2)), but not with O2px (there is no f(B1) from the H1s basis). This observation suggests that molecular orbitals in H2O will fall into the following groups:
b2
O2py
cðA1 Þ ¼ c1 fðO2sÞ þ c2 fðO2pz Þ þ c3 fðA1 Þ cðB1 Þ ¼ fðO2px Þ cðB2 Þ ¼ c4 fðO2py Þ þ c5 fðB2 Þ
b1
O2px
H1sA + H1s B
a1
b2 H1sA – H1s B Fig. 8.23 The symmetry classification of the oxygen atomic orbitals in H2O, a C2v molecule, and the two symmetry-adapted linear combinations of the H1s orbitals.
where we take z as the twofold rotation axis and x as perpendicular to the molecular plane. The secular determinant consists of three blocks, one being three-dimensional (involving the solution of a cubic equation for E), one being one-dimensional (involving only a trivial statement of the energy), and one being two-dimensional (and requiring the solution of a quadratic equation). In each case we have identified the symmetry species of the SALC by reference to the character table and have combined it with atomic orbitals of the same symmetry species to form a molecular orbital of the specified symmetry species. The molecular orbitals (not the SALCs) are labelled by lower case italic letters corresponding to the symmetry species, so in H2O we can expect the orbitals a1, b1, and b2. Each orbital of a particular symmetry species is then numbered sequentially in order of increasing energy, to give a notation such as 1a1, 2a1, and so on. The formal procedure for the construction of SALCs was explained in Section 5.12, where we saw that a character table is used to formulate a projection operator, and then that projection operator is applied to a member of the basis. The procedure is illustrated in the following example. Example 8.2 The construction of symmetry-adapted linear combinations
Construct the SALCs for the basis set given above for H2O. Method. Follow the method set out in Example 5.9. The point group is C2v
and h ¼ 4. Answer. The effect of the operations of the group on the basis is set out in the following table:
C2v
O2s
O2px
O2py
O2pz
H1sA
H1sB
E
O2s
O2px
O2py
O2pz
H1sA
H1sB
C2
O2s
O2px
O2py
O2pz
H1sB
H1sA
sv
O2s
O2px
O2py
O2pz
H1sB
H1sA
sv0
O2s
O2px
O2py
O2pz
H1sA
H1sB
268
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
For A1, d ¼ 1 and all w(R) ¼ þ1. Hence, column 1 gives O2s, column 2 and column 3 give 0, column 4 gives O2pz and column 5 gives 12(H1sA þ H1sB). This set of orbitals combine to give the molecular orbital c(A1) listed in the text. For A2, with characters (1,1, 1, 1), no column survives. For B1 with characters (1, 1,1, 1), column 2 gives O2px, and all other columns give 0. The B1 orbital is therefore a non-bonding orbital confined to the O atom, as no other orbitals present have net overlap with it (see Fig. 8.23). For B2, with characters (1, 1, 1,1), column 3 gives O2py and columns 5 or 6 give 1 2(H1sA H1sB). Hence, the B2 orbital has the form given in the text.
3a1
2b2
1b1 2a1
Comment. If there were d-orbitals available (as in H2S), the dz2 - and
dx2 y2-orbitals would contribute to A1 orbitals, dyz would contribute to B2, and dzx and dxy would be non-bonding and B1 and A2, respectively. 1b2
Self-test 8.2. Construct SALCs from the H1s orbitals of NH3.
1a1
Fig. 8.24 The molecular orbitals of
H2O at its equilibrium bond angle of 104 . a1
e
The energies of the orbitals and the values of the coefficients are found by solving the secular equations in the normal way. However, there is the added complication that the bond lengths and the bond angle must also be varied until the total energy of the molecule is a minimum, and that lowest energy arrangement of the atoms is accepted as the most stable state of the molecule. Alternatively, if the geometry of the molecule is known, then a single calculation may be carried out for that arrangement of nuclei. For H2O, for instance, the bond angle is 104 , and the molecular orbital energy level diagram is as shown in Fig. 8.24. As there are eight valence electrons to accommodate, the ground-state electron configuration of H2O is expected to be H2 O 1a21 1b22 2a21 1b21
a1
e
Fig. 8.25 The symmetry classification of the nitrogen atomic orbitals in NH3, a C3v molecule, and the three symmetry-adapted linear combinations of the H1s orbitals.
1
A1
The overall term symbol is calculated by multiplying together the characters of the occupied orbitals, and then identifying the overall symmetry species of the molecule from the character table. As all the orbitals are doubly occupied, and their characters are 1, the outcome is the set (1,1,1,1), which corresponds to A1. All electrons are paired, so S ¼ 0 and the multiplicity is 1. The same technique may be applied to ammonia, NH3, which belongs to the point group C3v. Now the minimum basis set, the basis set employing only the valence orbitals, consists of N2s, N2p, and three H1s, giving seven members in all. Without adopting symmetry arguments, we would expect to have to solve a 7 7 secular determinant. With symmetry taken into account, we would expect the problem to be reduced to a series of bite-sized determinants. Intuitively, we should expect the N2s- and N2pz-orbitals to belong to one symmetry species and N2px and N2py to belong to another. This separation can indeed be seen at a glance by looking at the C3v character table in Appendix 1, because an s-orbital and a pz-orbital on the central atom both span A1 whereas (px,py) jointly span E. The symmetry species of the three H atoms were established in Example 5.9, and we know that the SALCs, which there were called s1, s2, and s3, span A1 þ E. These points can be verified by reference to Fig. 8.25 or by reviewing the work that was done in Example 5.9.
8.9 CONJUGATED p-SYSTEMS
3a1
j
269
The complete 7 7 secular determinant for NH3 factorizes into a 3 3 determinant (for the A1 orbitals) and two 2 2 determinants (for the E orbitals). The molecular orbitals are therefore of the form cðA1 Þ ¼ c1 s1 þ c2 fðN2sÞ þ c3 fðN2pz Þ cðEÞ ¼ c1 0 s2 þ c2 0 fðN2px Þ and c1 00 s3 þ c2 00 fðN2py Þ
2e
2a1
1e
1a1
Fig. 8.26 The molecular orbitals of
NH3 at its equilibrium bond angle of 107 .
Fig. 8.27 The structure of the
p-orbital in ethene.
(The e-orbitals are distinguished by their reflection symmetry.) The solution of the secular determinant for the observed bond angle of 107 gives a set of energy levels shown in Fig. 8.26. There are eight electrons to accommodate, and so the configuration of the ground state is expected to be NH3
1a21 1e4 2a21
1
A1
The HOMO is the 2a1-orbital, which is largely a non-bonding orbital composed of N2s- and N2pz-orbitals: the electrons that occupy it therefore constitute a lone pair on the N atom.
8.9 Conjugated p-systems A special class of polyatomic molecules consists of those containing p-bonded atoms, particularly conjugated polyenes and arenes. They fall into a unique class because the orbitals with local s and p symmetry can be discussed separately. By ‘local’ symmetry we mean symmetry with respect to one internuclear axis rather than the global symmetry of the molecule. For global symmetry we have to classify orbitals according to the overall point group of the molecule, and the s,p designation is relevant only for linear species. However, if we focus on an individual A–B fragment of the molecule, then the orbitals do have a characteristic rotational symmetry about that axis, and they can be classified as locally s or p. One reason for the separate treatment of orbitals that can be classified locally as s and p is that the electrons in p-orbitals are typically less strongly bound than those in s-orbitals, so there is little interaction between the two types of orbital (recall the principles set out in Section 8.6). Another reason for the separation is that as p-orbitals are typically found in planar molecules, they have global symmetry properties (specifically, with respect to reflection in the plane) that distinguish them from s-orbitals, and therefore span different irreducible representations of the molecular point group. As a consequence, they can be discussed separately. The simplest organic p-system is the ethene molecule, CH2 ¼ CH2. The s-orbitals in ethene are molecular orbitals composed of various symmetryadapted linear combinations of C2s, C2px, C2py, and H1s orbitals; the p-orbitals are formed by overlap between C2pz orbitals where the z-axis lies perpendicular to the molecular plane (Fig. 8.27). This model immediately accounts for the torsional rigidity of the molecule, because (C2pz,C2pz)overlap is greatest when the molecule is planar. The p-orbital energies are found by solving a 2 2 secular determinant, and the solutions given in eqn 8.17 may be employed because the carbon–carbon fragment is homonuclear. When the p-system is conjugated, which means that the p-system extends over several neighbouring atoms, the simplest description of the bonding is in
270
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
terms of the Hu¨ckel approximation. This drastic approximation makes the following assumptions in the formulation of the secular determinant j H ES j: 1. All overlap integrals are set equal to zero: Srs ¼ drs. This is in fact a poor approximation, because actual overlap integrals are typically close to 0.2. Nevertheless, when the rule is relaxed, the energies are shifted in a simple way and their relative order is not greatly disturbed.
*
2. All diagonal matrix elements of the hamiltonian are ascribed the same value: Hrr ¼ a.
*
* 1 *
* *
*
* 2
The parameter a is negative. This approximation is reasonable for species that do not contain heteroatoms because all the conjugated atoms are electronically similar. Some justification comes from the Coulson–Rushbrooke theorem, which states that the charge density on all the carbon atoms is the same in alternant hydrocarbons. An alternant hydrocarbon is one in which the atoms can be divided into two groups by putting a star on alternate atoms and not having any neighbouring stars when the numbering is complete. Benzene (1) is alternant, azulene (2) is non-alternant. 3. All off-diagonal elements of the hamiltonian are set equal to zero except for those between neighbouring atoms, all of which are set equal to b. The parameter b is negative. It is the important parameter characteristic of Hu¨ckel theory, in so far as it governs the spacing of the molecular orbital energy levels. Example 8.3 The implementation of the Hu¨ckel approximation
Set up and solve the secular determinant for p-orbitals of the butadiene molecule in the Hu¨ckel approximation. – 1.618
Method. Construct the secular determinant by setting all diagonal elements
– 0.618
equal to a E and off-diagonal elements between neighbouring atoms equal to b; all other elements are zero. Set the secular determinant equal to zero, and solve the resulting quartic equation in x ¼ a E for x and hence E. Answer. The equation to solve is
+ 0.618
a E b 0 0
b aE b 0
0 0 b 0 ¼0 aE b b a E
On setting x ¼ a E and expanding the determinant, we obtain + 1.618
Fig. 8.28 The Hu¨ckel molecular
orbitals and their energies in butadiene (as viewed down the axis of the p-orbitals).
x4 3b2 x2 þ b4 ¼ 0 This quartic in x is a quadratic equation in y ¼ x2, so its roots can be found by elementary methods: 1=2 1=2 3 51=2 3 51=2 x¼ b x¼ b 2 2
8.9 CONJUGATED p-SYSTEMS
The secular determinant for butadiene is an example of a socalled ‘tridiagonal determinant’, in which the non-zero elements all lie along three neighbouring diagonal lines. From the theory of determinants, an N N tridiagonal determinant has the following roots: kp Ek ¼ a þ 2b cos N þ1 k ¼ 1, 2, . . . , N
j
271
We conclude that the energy levels are E ¼ a 1:618b a 0:618b as shown in Fig. 8.28. Self-test 8.3. Find the roots of the secular determinant for the p-orbitals of
square-planar cyclobutadiene.
[a 2b, a, a]
The worked example has shown how to calculate the molecular orbital energy levels in a simple case. The coefficients of the orbitals can be found by substituting these energies into the secular equations. However, in practice it is much easier to employ a computer: the roots we have found are the eigenvalues of the secular matrix H ES and the corresponding eigenfunctions of the matrix are the coefficients of the atomic orbitals that contribute to each molecular orbital. For example, the four molecular orbitals of butadiene are found in this way to be cð1pÞ ¼ 0:372fA þ 0:602fB þ 0:602fC þ 0:372fD cð2pÞ ¼ 0:602fA 0:372fB þ 0:372fC þ 0:602fD cð3pÞ ¼ 0:602fA þ 0:372fB þ 0:372fC 0:602fD cð4pÞ ¼ 0:372fA 0:602fB þ 0:602fC 0:372fD
Fig. 8.29 The contributions of the
p-orbitals to each p-orbital match the amplitude of a sine wave (the wavefunction for a particle in a
where fJ is a 2pz-orbital on atom J. The composition of these molecular orbitals is independent of the values of a and b (provided b 6¼ 0). Notice that the energy of the orbital increases with the number of nodes, and that the amplitude of each coefficient follows a sine wave fitted to the length of the molecule (Fig. 8.29). The ground-state configuration of the molecule is 1p22p2, which corresponds to a total p-electron energy of 4a þ 2(5)1/2b. The energy of a single unconjugated (bonding) p-orbital is a þ b (see eqn 8.16 and ignore overlap), and so if the molecule were described as having two unconjugated p-bonds, its total p-electron energy would have been 4a þ 4b. The difference, which in this case is 2(51/2 2)b ¼ 0.472b, is called the delocalization energy of the molecule. The delocalization energy is independent of a within the Hu¨ckel approximation largely because all atoms are equivalent and the total electron density on them is the same regardless of the extent of delocalization of the orbitals. Remember that b is negative. A very approximate order-ofmagnitude value is b ¼ 0.75 eV (72 kJ mol1). The Hu¨ckel procedure leads to secular determinants of large dimension. However, they may often be factorized into more manageable dimensions by making use of the symmetry of the system beyond the simple mirror plane that enables the p-system to be distinguished from the s-system. This additional factorization follows from the usual arguments about the hamiltonian having no non-zero elements between linear combinations of orbitals that belong to different symmetry species of the molecular point group. For benzene, for instance, the 6 6 determinant can be simplified considerably by making use of the D6h symmetry of the molecule. In fact, because every 2pz-orbital changes sign under reflection in the molecular plane, we lose no information by using the C6v subgroup of the molecule. The procedure involves treating the C atoms
272
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
− 2 b2 g
− e2u
0
+ e1g
+ 2 a2u
Fig. 8.30 The Hu¨ckel molecular
orbitals and their energies in benzene.
as the peripheral atoms of a molecule, and setting up SALCs of their 2pz-orbitals; however as there is no ‘central’ atom, these SALCs are in this instance the actual p molecular orbitals of the molecule. The projection operator technique described in Section 5.12 leads to the following linear combinations (labelled according to the symmetry species of the group D6h): 1=2 1 cða2u Þ ¼ ðpA þ pB þ pC þ pD þ pE þ pF Þ 6 1=2 1 cðe1g Þ ¼ ðaÞ ð2pA þ pB pc 2pD pE þ pF Þ and 12 1 ðbÞ ðpB þ pC pE pF Þ 2 1=2 1 ð2pA pB pC þ 2pD pE pF Þ and cðe2u Þ ¼ ðaÞ 12 1 ðbÞ ðpB pC þ pE pF Þ 2 1=2 1 ðpA pB þ pC pD þ pE pF Þ cðb2g Þ ¼ 6 These orbitals are sketched in Fig. 8.30. Note that the form of the orbitals is determined solely by the symmetry of the molecule and makes no reference to the values of a or b. As we show in the following example, the energy levels are Eðe1g Þ ¼ a þ b Eðe2u Þ ¼ a b Eðb2g Þ ¼ a 2b Eða2u Þ ¼ a þ 2b As we have already remarked, b is negative, so the orbitals lie in the order shown in the illustration. Example 8.4 The energy levels of the benzene molecule
Determine the p-electron energy levels of the benzene molecule by using the Hu¨ckel approximation. Method. The molecular orbitals are specified above. We need form secular
determinants for each orbital species separately as the hamiltonian has no offdiagonal elements between orbitals of different symmetry species. Use the Hu¨ckel rules for writing the matrix elements after expanding the Hrs in terms of the linear combinations of 2pz-orbitals. The orbitals that span one-dimensional irreducible representations will give simple 1 1 determinants, which are trivial to solve. The orbitals that span two-dimensional irreducible representations will give 2 2 determinants, which will lead to quadratic equations. However, because the e-orbitals of each set have different reflection symmetry, they too give diagonal determinants, so the roots can be found trivially. Answer. The matrix elements we require are as follows:
ha2u jHja2u i ¼ 16hpA þ þ pF jHjpA þ þ pF i ¼ a þ 2b hb2g jHjb2g i ¼ 16hpA pF jHjpA pF i ¼ a 2b he1g ðaÞjHje1g ðaÞi ¼ a þ b he1g ðbÞjHje1g ðbÞi ¼ a þ b he2u ðaÞjHje2u ðaÞi ¼ a b he1g ðaÞjHje1g ðbÞi ¼ 0
he2u ðbÞjHje2u ðbÞi ¼ a b he2u ðaÞjHje2u ðbÞi ¼ 0
8.9 CONJUGATED p-SYSTEMS
j
273
The resulting energies are those quoted in the text and displayed in Fig. 8.30. Comment. For a cyclic polyene of formula CNHN containing N carbon atoms
in the ring, the general solution of the secular determinant yields the energy levels 2kp Ek ¼ a þ 2b cos N where k ¼ 0, 1, 2, . . . , (N 1)/2 for odd N and k ¼ 0, 1, 2, . . . , (N 2)/2, N/2 for even N. This result is the basis of a simple graphical mnemonic for relating the energy levels of a cyclic polyene to its shape. As shown in Fig. 8.31, the pattern of energy levels mirrors the locations of the carbon atoms (which lie at locations given by cos(2kp/N) around the ring of an N-atom polyene). Self-test 8.4. Use the C2v subgroup of naphthalene to find the p-electron
molecular orbital energy levels within the Hu¨ckel approximation.
The ground-state electron configuration of benzene is C6 H6
a22u e1g4
1
A1g
and the delocalization energy is Edeloc ¼ ð6a þ 8bÞ 6ða þ bÞ ¼ 2b The six electrons just complete the molecular orbitals with net bonding effect, leaving unfilled the orbitals with net antibonding character, which is a characteristic configuration for aromatic molecules. To some extent this configuration echoes the configuration of N2, and both molecules have a pronounced chemical inactivity. Another feature of the energy levels of benzene is that the array of levels is symmetrical: to every bonding level there corresponds an antibonding level. This symmetry is a characteristic feature of
Fig. 8.31 The pattern of energy
levels in cyclic polyenes mirrors the locations of the carbon atoms in the ring.
274
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
alternant hydrocarbons and can be traced to the topological character of the molecules. Indeed, many of the results of Hu¨ckel theory can be established on the basis of ‘graph theory’, the branch of topology concerned with the properties of networks. One particular result of this kind of analysis is the justification of the ‘(4n þ 2)-rule’ for the anticipation of aromatic character, where n is the number of p-electrons. As we have stressed, Hu¨ckel theory, which virtually hijacks the disagreeable integrals that appear in a full treatment, is only the most primitive stage of discussing p-electron molecules.8 The modern, far more reliable numerical approaches are described in Chapter 9.
8.10 Ligand field theory The success of Hu¨ckel theory is rooted in the fact that the orbitals themselves are determined by the symmetry of the system. These symmetry-determined orbitals are then put into an order of energies, essentially by counting the number and noting the importance of their nodes. The energy differences between the orbitals are typically so large that the coarseness of this procedure does not unduly misrepresent their order. A similar situation occurs in the complexes of d-metal ions. These complexes consist of a central metal ion surrounded by a three-dimensional array of ligands. The composition of the orbitals of the complex is largely determined by the symmetry of the environment, and a single parameter can be used to give a rough indication of the order of the energies of the molecular orbitals of the complex. Ligand field theory is a kind of three-dimensional version of Hu¨ckel theory, in which symmetry plays a central role, and in which structural, spectroscopic, magnetic, and thermodynamic properties are parametrized in terms of the ligand field splitting parameter, D. We denote the central metal ion by M and assume that it has the configuration dn. The ligands are denoted L, and we confine attention to ML6 octahedral complexes with Oh symmetry. The orbitals of the ligands are denoted l. In particular, we suppose that each ligand i supplies an orbital li(s) that has local s symmetry with respect to the M L bond. Thus, in ligand field theory, each Lewis-base ligand is simulated by a single orbital that supplies two electrons. Later we shall allow for the possibility that the ligands can supply electrons from or accept electrons into their p-orbitals, and denote the ðpÞ latter by li . The first step in ligand field theory is to set up symmetry-adapted linear combinations and to identify the symmetry species of the d-orbitals on the central metal ion. A glance at the Oh character table in Appendix 1 shows that in an octahedral environment, two of the d-orbitals (dz2 and dx2 y2 ) span Eg and the remaining three (dxy, dyz, and dzx) span T2g. The standard techniques of group theory show that the ligand s-orbitals span A1g þ Eg þ T1u
.......................................................................................................
8. In a more sophisticated version of Hu¨ckel theory, called extended Hu¨ckel theory, overlap integrals are not neglected. See the Further reading section for details.
8.10 LIGAND FIELD THEORY
a1g
3
1
5 2 4
eg
6
t1u Fig. 8.32 A depiction of the
Metal
Complex t1u
Ligands
a1g
Antibonding
symmetry-adapted linear combination of ligand atomic orbitals in an octahedral complex.
4p 4s
3d
Nonbonding
eg ∆
t2g
eg t1u
t1u a1g
Bonding
a1g eg
Fig. 8.33 The molecular orbital
energy level diagram for an octahedral complex. For accounting purposes, the ligand electrons occupy all the bonding orbitals; the electrons supplied by the metal atom occupy the orbitals in the box.
j
275
in Oh, and projection operator techniques give the following explicit forms of the corresponding SALCs: 1=2 1 ðsÞ ðsÞ ðsÞ ðsÞ ðsÞ ðsÞ l1 þ l2 þ l3 þ l4 þ l5 þ l6 cðA1g Þ ¼ 6 1=2 1 ðsÞ ðsÞ ðsÞ ðsÞ ðsÞ ðsÞ 2l5 þ 2l6 l1 l2 l3 l4 cðEg Þ ¼ ðaÞ and 12 1 ðsÞ ðsÞ ðsÞ ðsÞ ðbÞ l1 þ l2 l3 l4 2 1=2 1=2 1 1 ðsÞ ðsÞ ðsÞ ðsÞ cðT1u Þ ¼ ðaÞ l1 l2 , ðbÞ l3 l4 , and 2 2 1=2 1 ðsÞ ðsÞ ðcÞ l5 l6 2 These SALCs are illustrated in Fig. 8.32. Note that there is no T2g combination. These combinations differ in energy slightly when we take into account overlap between ligand orbitals (as distinct from the (M,L) overlap that is predominantly responsible for bonding). Now we form molecular orbitals as linear combinations of the SALCs and the d-orbitals of the same symmetry species, for only these combinations have non-zero net overlap. It is apparent from Fig. 8.32 that dz2 has non-zero overlap with c(Eg, a) but not with c(Eg, b); the opposite is true for dx2 y2 . cða1g Þ ¼ cðA1g Þ cðeg Þ ¼ ðaÞ c1 fðdz2 Þ þ c2 cðEg , aÞ and ðbÞ c01 fðdx2 y2 Þ þ c02 cðEg , bÞ cðt1u Þ ¼ cðT1u Þ cðt2g Þ ¼ ðaÞ dxy , ðbÞ dyz , ðcÞ dzx We have not included overlap with s- and p-orbitals: they transform as A1g and T1u, respectively, and so would combine with the SALCs of those symmetry species. Within the d-orbital-only approximation, we see that we can expect an array of energy levels like that shown in Fig. 8.33. The bonding eg combination is largely confined to the ligands (the lower energy orbitals) and the antibonding combination is largely confined to the metal ion. The a1g and t1u combinations labelled ‘bonding’ in Fig. 8.33 are confined almost entirely to the ligands. The t2g orbitals are non-bonding atomic orbitals on the metal ion. There are 12 electrons to accommodate that are supplied by the ligands (two from each Lewis base), and n electrons supplied by the metal ion. Of these 12 þ n electrons, 12 fill the two bonding eg and four ‘bonding’ a1g and t1u orbitals: these electrons are largely confined to the ligands. Up to six of the remaining n electrons are free to occupy the three t2g orbitals on the metal ion and the remainder will occupy the antibonding eg combination, which is largely confined to the metal ion too. However, at this point there is a complication. The ground-state electron configuration of the complex is the configuration that corresponds to the lowest total energy. When the separation between t2g and the antibonding eg orbitals is small, it may be advantageous to occupy the latter orbital before completely filling the former, because then the electrons occupy spatially
276
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
Strong field
Weak field
4
5
6
7
Fig. 8.34 The high- and low-spin
arrangements that arise from weak and strong ligand fields for complexes with four to seven d-electrons.
distinct regions and may do so with parallel spins and so benefit from spin correlation. The crucial quantity is the ligand field splitting parameter, D, the energy separation between eg and t2g. If this splitting is large, then a d4 4 complex, for instance, will adopt the configuration t2g with one orbital doubly occupied. However, if the splitting is small, then it may be energetically 3 1 eg with all four advantageous for the complex to adopt the configuration t2g electrons in separate orbitals with parallel spins. It follows from this discussion that we should distinguish between the following two cases: 1. The strong-field case, in which the ligand field splitting parameter is large and it is energetically favourable to occupy the t2g orbitals first. 2. The weak-field case, in which the ligand field splitting parameter is small and it is energetically favourable to occupy the eg orbitals before the t2g orbitals are completely filled. As an indication of magnitudes, the ligand field splitting parameter in [Cr(CN)6]3 (a d3 strong-field case) is 26 600 cm1 (3.30 eV) whereas that in [Cr(OH2)6]3þ (a d3 weak-field case) is 17 400 cm1 (2.16 eV). The second category is sometimes further divided into ‘weak field’ itself and very weak field, according to whether the ligand field splitting parameter is stronger or weaker, respectively, than the spin–orbit interaction. The very weak field case is applicable to the f-block elements in which the f-electrons are embedded deeply in the atom and experience the surrounding ligands only very weakly. We shall not consider it further. The ambiguity in ground-state configuration is found for d4, d5, d6, and 4 5 6 6 1 7 d complexes. When the ligand field is so large that a t2g , t2g , t2g , t2g eg configuration is adopted, the spins need to pair. As a result, such complexes are classified as low-spin complexes. As may be verified from Fig. 8.34, they have 2, 1, 0, and 1 unpaired spins, respectively. When the ligand field 3 1 3 2 4 2 5 2 eg , t2g eg , t2g eg , and t2g eg , is weak, the complexes can be expected to be t2g with 4, 5, 4, and 3 unpaired spins. Such complexes are classified as high-spin complexes. Because the number of unpaired electrons is responsible for the magnetic properties of complexes, we see that modification of the ligands and consequently the size of D may influence the magnetic properties of the species.
8.11 Further aspects of ligand field theory There are three aspects of ligand field theory that need to be touched on here. In the first place, we need to be aware that the weak-field case can be quite n n0 eg configurations are so strongly perturbed tricky to handle because the t2g by electron–electron interactions. We shall illustrate the difference between high-field and low-field cases by considering a d2 configuration. In a free ion, a d2 configuration, as in Ti2 þ and V3 þ , can give rise to the terms 1G, 3F, 1D, 3P, and 1S. From Hund’s rules, we can expect the 3F term to lie lowest in energy, with perhaps the 3P next above it. When the ligand field is weak, we can think of the formation of molecular orbitals as a small perturbation on the free-ion levels. To determine the effect of this
8.11 FURTHER ASPECTS OF LIGAND FIELD THEORY d
23
P
eg A 2g 2 3
eg t2g T1g
Energy
1
1 3
eg1t2g1 3 T2g
d
23
F
t2g T1g 2 3
Strong field
Fig. 8.35 A Tanabe–Sugano
type of correlation diagram for the states of an octahedral two-electron complex.
277
perturbation, we consider the effect of the reduction in symmetry from R3 (the full rotation group in three dimensions, typical of an atom) to Oh and identify the symmetry species that 3F and 3P become in the octahedral environment. The technique required was described in Section 5.19 and illustrated in Example 5.14: 3
P ! 3T1g
3
F ! 3A2g þ 3 T1g þ 3 T2g
The separation of the terms stemming from 3F increases as the perturbation becomes stronger (Fig. 8.35). In the strong-field case we can discuss the configurations in terms of occupation of the t2g and eg orbitals, and we can have 2 t2g
Weak field
j
3
T1g
1 1 t2g eg
3
T1g þ 3 T2g
e2g
3
A2g
(we have retained only the triplet terms). The order of energies can be anticipated by referring to Fig. 8.33, and they are shown on the right of Fig. 8.35. At this point, we can construct the correlation diagram by connecting states of the same symmetry but allowing for the non-crossing rule. The diagram in Fig. 8.35 is in fact a part of a Tanabe–Sugano diagram for the correlation of strong and weak field states of a complex. The actual state of a complex corresponds to an intermediate stage of the diagram, and the location can be determined by fitting the observed spectroscopic transitions to the energy levels. In practice, a Tanabe–Sugano diagram is expressed in terms of quantities that parametrize the strengths of the electron–electron repulsion and the ligand field splitting parameter, so these quantities can be determined. More information on the procedures will be found in the books referred to in the Further reading section for this chapter. The second feature we need to mention concerns deviations from octahedral symmetry that arise spontaneously in certain complexes such as many of the hexa-coordinate copper(II) complexes. Their occurrence is summarized by the Jahn–Teller theorem:
dz 2
Energy
In any non-linear system, there exists a vibrational mode that removes the degeneracy of an orbitally degenerate state.
dx 2 – y 2
Distortion
Fig. 8.36 The effect of the
distortions envisaged in the Jahn–Teller effect.
The theorem can be illustrated by considering a d9 octahedral complex, such 6 3 eg ; it as [Cu(OH2)6]2þ. The ground-state configuration is expected to be t2g 10 is orbitally degenerate because the ‘hole’ in the d configuration can occupy either dz2 or dx2 y2 . If the complex were to distort so that it lengthened along a C4 axis, then the degeneracy of the antibonding eg orbital would be removed by the change in overlap. We would expect the antibonding character of the L orbital formed by overlap with the dz2 -orbital to be reduced as the M length increases, so this molecular orbital will become lower in energy (Fig. 8.36). Alternatively, if the M–L length were to shorten, the same eg orbital would become higher in energy. In either case, the complex will remain in the distorted shape, because it then has a lower energy than in the undistorted, regular octahedral shape. A final detail concerns the role of p-bonding between the metal ion and the ligands. We shall suppose that on each ligand there are two orbitals
278
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
1
6 z
4
y 3 x 2
5
with local p symmetry with respect to the M–L axis. They span T1u þ T2u þ T1g þ T2g (Fig. 8.37), and only the last can have net overlap with the t2g orbitals of the ion. The explicit structures of the SALCs of these orbitals are as follows: cðt1u Þ ¼ ðaÞ 12ðp3x þ p4x þ p5x þ p6x Þ, ðbÞ 12ðp1y þ p2y þ p5y þ p6y Þ, ðcÞ 12ðp1z þ p2z þ p3z þ p4z Þ cðt2g Þ ¼ ðaÞ 12ðp5x p6x þ p1z p2z Þ, ðbÞ 12ðp5y p6y p3z þ p4z Þ,
t1u
ðcÞ 12ðp1y p3x p2y þ p4x Þ cðt1g Þ ¼ ðaÞ 12ðp6x p5x þ p1z p2z Þ, ðbÞ 12ðp5y p6y þ p3z p4z Þ, ðcÞ 12ðp1y p2y þ p3x p4x Þ cðt2u Þ ¼ ðaÞ 12ðp5y þ p6y p1y p2y Þ, ðbÞ 12ðp1z þ p2z p3z p4z Þ,
t2g
t1g
t2u Fig. 8.37 A representation of
the symmetry-adapted linear combinations of ligand p-orbitals. Each arrow can be regarded as indicating a p-orbital, the head indicating the positive lobe.
ðcÞ 12ðp5x þ p6x p3x p4x Þ Two cases may be distinguished, and are illustrated in Fig. 8.38. In the first, the ligand p-orbitals are full (the ligands act as p-donors). In this case, the ligand field splitting parameter is reduced by the formation of (M,L) p-orbitals because the original non-bonding t2g orbitals become slightly antibonding. In the second case, in which the p-orbitals of the ligands are initially empty (so the ligands act as p-acceptors), the t2g orbitals are made slightly bonding, with the result that the ligand-field splitting parameter is increased. The correlation of the value of D with the identity of the ligand and the metal ion depends critically on the ability of the species to form p-orbitals. Moreover, the stability of complexes such as those formed by CO (a p-acceptor ligand, recall the discussion in Section 8.7), including Ni(CO)4 and [Fe4(CO)13]2, can also be traced to the involvement of p-orbitals.
The band theory of solids The electronic structures of solids can be regarded as an extension of molecular orbital theory to aggregates consisting of virtually infinite numbers of atoms. However, there are certain features that are unique to solids, particularly the formation of continuous bands of energy levels instead of discrete levels, and the role of the translational symmetry of the lattice. t2g
eg
∆
Fig. 8.38 The effect of p-bonding on
the ligand field splitting in an octahedral complex: (a) occupied ligand p-orbitals (a p-donor ligand) and (b) unoccupied ligand p-orbitals (a p-acceptor ligand).
d
t2g eg
d
eg ∆
t2g
(a)
t2g
(b)
eg
8.12 THE TIGHT-BINDING APPROXIMATION
j
279
There are in fact two starting points for the discussion of solids. One is the particle-in-a-box wavefunctions described in Chapter 2. The other is the discussion of conjugated molecules presented earlier in this chapter. We shall give a brief introduction to both and see how one may be correlated with the other. We shall confine our attention to one-dimensional solids because they are so much simpler to treat. However, such solids do not show all the properties of a three-dimensional solid, and this material must be regarded as no more than introductory. Fig. 8.39 The string of s-orbitals used to discuss the formation of bands in a one-dimensional solid.
–2
The tight-binding approximation treats a solid as an extended molecule, and takes as its starting point orbitals that are confined to individual atoms (hence the name of the approach). Then molecular orbitals are formed that spread throughout the solid. The simplest approach of all is to adopt the Hu¨ckel approximation and to consider a line of N atoms, each of which has one valence s-orbital that can overlap only its two immediate neighbours (Fig. 8.39). As usual, the wavefunctions will be X cr fr ð8:36Þ c¼
Energy
r
0
2
8.12 The tight-binding approximation
1 2 3 4 5 6 7 8 9 101112 ∞
Fig. 8.40 The formation of molecular orbitals from a chain of N atomic orbitals. Note that the separation of the most bonding and most antibonding orbitals remains finite and that the density of orbitals is greatest at the edges of the band.
where the index r runs over determinant has the form a E b 0 0 b a E b 0 0 b a E b .. .. .. .. . . . .
all the atoms in the line. The N N secular ¼ 0 .. .
This determinant is tridiagonal (see Example 8.3), so we can write down the roots immediately: kp Ek ¼ a þ 2b cos k ¼ 1, 2, . . . , N ð8:37Þ Nþ1 As N ! 1 the energy separation between neighbouring levels approaches zero but the width of the band remains finite (Fig. 8.40): ð8:38Þ lim ðE1 EN Þ ¼ 4b N!1
The lowest energy corresponds to a fully bonding linear combination of atomic orbitals and the highest energy corresponds to a molecular orbital that has a node between each pair of neighbouring atoms. The molecular orbitals of intermediate energy have k 1 nodes distributed along the chain of atoms. Example 8.5 The density of states of a one-dimensional solid
Inspection of the diagram in Fig. 8.40 indicates that the density of states, r(E), the number of states in an energy range divided by the width of the range, increases towards the edges of the bands. Confirm this conclusion analytically.
j
280
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
Method. First, we need to define the density of states analytically. Each value
of k denotes a single state. If the quantum number k changes from k to k þ Dk as the energy changes from E to E þ DE, there are Dk states in the energy range DE, so the density of states at the energy E is r(E) ¼ Dk/DE. When the states are packed together so closely that to a good approximation they form a continuum, we can replace the finite quantities by infinitesimals and write r(E) ¼ dk/dE. It is simpler to express the energy as e ¼ E a. Answer. We write the energy expression as
kp Nþ1 from which it follows that Nþ1 e k¼ arccos p 2b e ¼ 2b cos
We have used the result that d arccos ax dx ¼
a
ð1 a2 x2 Þ
1=2
The density of states is therefore dk ðN þ 1Þ=2bp ¼ n rðeÞ ¼ o1=2 de 1 ðe=2bÞ2
8
This function (noting that b < 0) becomes infinite at the edges of the band, where e ¼ 2b (Fig. 8.41).
2π| | ( ) / (N +1)
6
Comment. The graphical mnemonic in Example 8.4 illuminates this conclu-
sion, for when N is very large there is little difference between a line of atoms and one with the ends joined to form a circle. There are more points captured by slices near the top and bottom of the circle than at its mid-point. The increase of the density of states towards the edge of the band is uncharacteristic of higher-dimensional solids. In them, the density is highest towards the centre of the band, and is least at the edges. This difference arises from the possibility of degeneracies in dimensions greater than 1.
4
2
0
–1
/2
1
Self-test 8.5. Derive an expression for the mean energy of a band that is
half full.
Fig. 8.41 The density of states in
a band formed from an infinite chain of atomic orbitals.
The band of orbitals we have constructed is called an s-band because it is formed by the linear combination of s-orbitals. If the atoms have valence p-orbitals too, then a similar superposition can take place, with the formation of a p-band. In a typical solid, the energy separation of the s- and p-orbitals of the free atoms will be quite large, and as a result the two bands will not overlap. The orbital structure of the solid will therefore consist of two (or more) bands separated by a band gap, a region of energy to which no orbitals belong. If each atom provides one electron, the s-band will be half full. The band is then known as a conduction band because the electrons in the highest filled orbitals can travel through the solid in response to the application of electric fields. If, however, each atom provides two electrons, then the s-band will be full. It is then called a valence band. Its uppermost electrons are separated by a substantial energy gap from the p-band, and so they are not mobile. A metallic conductor is a substance with an electric conductivity that decreases as the temperature is increased. Such materials have incomplete
8.13 THE KRONIG–PENNEY MODEL
j
281
conduction bands and the decrease in conductivity arises from the increased scattering of the mobile electrons by lattice vibrations. A semiconductor is a solid with an electric conductivity that increases as the temperature is increased. Such materials have a full valence band separated by a small gap from an empty conduction band. Their conductivity arises from the excitation of electrons from the valence band into the conduction band, and the number of electrons so promoted increases with temperature. If the band gap is large compared with kT, so that the conductivity is very low, then the material is termed an insulator. The artificial manipulation of the properties of conduction and valence bands by the insertion of foreign atoms is the basis of the semiconductor industry, and further information can be found in the references given in Further reading.
8.13 The Kronig–Penney model We now turn to a seemingly entirely different attack on the same problem. This approach will echo the material in Chapter 2, and—most surprisingly— results in a description of solids that, despite the entirely different starting point, mirrors the molecular-orbital approach. The origin of that similarity is another manifestation of the power of symmetry, in this case translational symmetry, in determining the general structure of energy levels regardless of the details of physical interactions. If we were to disregard the variation in the potential energy of an electron as it travels through the lattice, the solutions of the Schro¨dinger equation would be those of a free particle, and we would write9 k2 h2 ð8:39Þ 2me Note how the energy varies quadratically with the wavevector k. In an actual solid, the potential energy varies periodically, and the Schro¨dinger equation is ck ðxÞ ¼ eikx
Ek ¼
2 d2 c h þ VðxÞc ¼ Ec 2me dx2
ð8:40Þ
with V(x þ a) ¼ V(x), where a is the spacing of the lattice points. According to the Bloch theorem, the solutions of the Schro¨dinger equation for a periodic potential of this kind have the form ck ðxÞ ¼ uk ðxÞeikx V
–b 0 a
x
Fig. 8.42 The potential energy
of an electron in the Kronig–Penney model.
uk ðx þ aÞ ¼ uk ðxÞ
ð8:41Þ
The periodic functions uk(x) are called Bloch functions. Substitution of the function uk(x)eikx into the Schro¨dinger equation leads to the following equation for the Bloch functions: 2me 2 uk ¼ 0 u00k þ 2iku0k fVðxÞ Eg þ k ð8:42Þ h2 where u 0 ¼ du/dx and u00 ¼ d2u/dx2. We shall establish the form of the Bloch functions in the particular case of a periodic potential energy like that shown in Fig. 8.42, which is called the
.......................................................................................................
9. We consider here only linear momentum in the positive x direction and therefore ignore the free-particle solutions eikx.
282
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
Kronig–Penney model. It is plainly a great simplification of the true potential energy, but it establishes certain important features that are found in practice. There are two types of region, one in which the potential is zero and the other in which it has the constant value V. We shall consider solutions for which E < V and shortly simplify the problem still further by letting V ! 1 and b ! 0 in such a way that Vb (the area of the rectangular region of nonzero potential energy) remains constant. It will be convenient to introduce the two real parameters a2 ¼
2me E h
2
b2 ¼
2me ðV EÞ 2 h
ð8:43Þ
and then to write the equations for the two regions as ðaÞ V ¼ 0 :
uk 00 þ 2ikuk 0 þ ða2 k2 Þuk ¼ 0
ðbÞ V 6¼ 0 :
uk 00 þ 2ikuk 0 ðb2 þ k2 Þuk ¼ 0
The solutions are subject to the requirement that the wavefunctions and their first derivatives are continuous at the interfaces between the regions. As may be verified by substitution, the solutions of the two differential equations have the form ðaÞ uk ðxÞ ¼ AeiðakÞx þ BeiðaþkÞx ðbÞ uk ðxÞ ¼ CeðbikÞx þ DeðbþikÞx The conditions of continuity of u and u 0 at the two boundaries of each zone, namely uk(a) ¼ uk(b) and uk 0 (a) ¼ uk 0 (b), lead to the following four equations: AþBCD¼0 AeiðakÞa þ Beiða þ kÞa CeðbikÞb DeðbþikÞb ¼ 0 iða kÞA iða þ kÞB ðb ikÞC þ ðb þ ikÞD ¼ 0 iða kÞAeiðakÞa iða þ kÞBeiðaþkÞa ðb ikÞCeðbikÞb þ ðb þ ikÞDeðbþikÞb ¼ 0 For these four simultaneous equations to have a solution, the determinant of the coefficients must be zero: 1 1 1 1 iðakÞa iðaþkÞa ðbikÞb ðbþikÞb e e e e ¼ 0 iða kÞ iða þ kÞ ðb ikÞ ðb þ ikÞ iða kÞeiðakÞa iða þ kÞeiðaþkÞa ðb ikÞeðbikÞb ðb þ ikÞeðbþikÞb ð8:44Þ This rather horrendous determinant reduces (as can best be shown by use of symbolic algebra software, but patience and a pencil also work) to the condition ! b2 a2 sinh bb sin aa þ cosh bb cos aa ¼ cos kða þ bÞ ð8:45Þ 2ab
8.13 THE KRONIG–PENNEY MODEL
j
283
sinaa + cos a
When we introduce the simplifying conditions V ! 1, b ! 0, Vb ¼ constant, eqn 8.45 simplifies to g
1
0 –1 –10
0
10
a
Fig. 8.43 The solution of the
equation for the energy levels of the Kronig–Penney model. The only permitted solutions are those that correspond to the regions in which the curve lies between þ1 and 1.
0
sin aa þ cos aa ¼ cos ka aa
g¼
me Vba
ð8:46Þ
a h2
This equation is still transcendental, but we can identify its implications by plotting the left-hand side against aa. The left-hand side depends on the value of g, which is a measure of the height and width of the barrier between neighbouring wells, and one such graph is shown in Fig. 8.43, where we have used g ¼ 32p. The essential point can now be made clear. Because the right-hand side of eqn 8.46 lies between 1 and þ1, only certain values of aa (that is, only certain values of E, because a / E1/2) give rise to solutions. Where the left-hand side lies outside the range 1 to þ1, there are no solutions. It follows that the solutions of the Schro¨dinger equation for a periodic potential correspond to a series of allowed bands separated by gaps. Moreover, as can be seen from Figs 8.43 and 8.44, the widths of the allowed bands increase with increasing energy. A final important point can be seen by comparing the diagrams in Fig. 8.44, which show the effect of changing g, the depth of the potential wells. As can be seen, as the depth increases, the allowed regions become narrower and converge on those of a particle in a square well.
∞
Energy
Example 8.6 The asymptotic behaviour of a periodic solid
Show that for infinitely deep wells, the energy spectrum of a periodic solid becomes that of a collection of independent wells. Method. When examining an equation for its asymptotic solutions, identify
the terms that dominate the others as the selected parameter becomes infinite, and retain only them. Find the solution of the remaining terms.
Fig. 8.44 The formation of bands as
a function of the parameter g in the Kronig–Penney model. The limit g ¼ 0 corresponds to the absence of barriers, and there is no discrete band structure (the levels are those of a free particle, so there is one infinitely wide band). The other limit, g ¼ 1, corresponds to a series of independent infinitely deep square wells, and each energy level corresponds to those of a particle in a box of width a.
Answer. As g ! 1, the first term in eqn 8.46 dominates the other two and the equation becomes sin aa ’0 g aa This equation has solutions only for aa ¼ np with n ¼ 1, 2, . . . . It follows that the allowed energies are
En ¼
2 a2 h n2 h2 ¼ 2me 8me a2
exactly as for a particle in a single box. Comment. The electron cannot tunnel between neighbouring boxes when the
wells are infinitely deep. Self-test 8.6. Derive an expression for the width of the allowed band. [2 arctan (g/aa)]
284
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
Equation 8.46 can be solved numerically for a as a function of k, and hence the variation of the energy with k can be determined. The variation for g ¼ 32p is shown in Fig. 8.45. The discontinuities occur at k ¼ np=a
5
Fig. 8.45 The band structure
Energy
represented as a plot of energy against the parameter k.
–5 –4 –3 –2 –1 0 1 2 3 4 5 ka/π Fig. 8.46 In this depiction of the
band structure, the curves illustrated in the previous diagram have all been transferred into the central zone to give a more compact representation.
ð8:47Þ
8.14 Brillouin zones
Energy –5 –4 –3 –2 –1 0 1 2 3 4 ka/π
n ¼ 1, 2, . . .
Each region between the discontinuities in eqn 8.47 is called a Brillouin zone. The discontinuities occur at the edges of the Brillouin zones, and towards the centres of the zones the variation of E with k is parabolic, exactly as in a free particle model. This observation, together with eqn 8.47, is a clue to the origin of the existence of band gaps, because they occur where the periodicity of the lattice matches the periodicity of the wavefunctions. In the centres of the zones, there is no match between the two, and the combinations cos kx and sin kx of the complex wave eikx are degenerate because, averaged over the lattice, each combination samples favourable regions of the potential equally. However, when the periods match, the cosine function (for instance) has maximum probability in the wells throughout the solid and the sine function has nodes in the wells everywhere. The two combinations are now no longer degenerate, and the perturbation caused by the lattice has driven them apart (just as in famous Fig. 6.1). It should be noted that k in the right-hand side of eqn 8.46 can be changed to k 2pn/a without changing the value of the right-hand side. So we can adjust the value of k by this amount at will, yet still obtain the same energies. In the reduced wavevector representation, k is modified by a different amount in each Brillouin zone to bring its value into the range p/a k p/a. This reduction has the effect of compressing Fig. 8.45 into the form shown in Fig. 8.46, where all the values of k lie in a range of width 2p/a. Finally, we impose a further constraint on the wavefunctions. When there are many atoms (wells) in the lattice, there is little error introduced if we assume that the ends of the lattice can be brought round into a circle and joined. This procedure preserves the translational symmetry of the system throughout its length rather than introducing awkward end effects. (If we were interested in the surface states of metals, such a procedure would be invalid, of course.) The circularity of the system implies that the wavefunctions must satisfy cyclic boundary conditions (Section 3.1), and that c(x þ L) ¼ c(x) where L ¼ Na, N being the number of atoms in the ring. In terms of the Bloch functions, this condition is uk ðx þ LÞeikðxþLÞ ¼ uk ðxÞeikx However, because uk(x þ L) ¼ uk(x) (as one location is an integral number of lattice periods a away from the other), this condition is equivalent to 2pn 2pn ¼ n ¼ 0, 1, 2, . . . k¼ L Na We have seen, however, that an entire zone is expressed by values of k that lie within a length 2p/a. It follows that j n j cannot exceed 12N, for otherwise k would lie outside the range. Therefore, the number of spatial states in any band of the system is N. Another way of accepting the validity of this result is
PROBLEMS
j
285
to consider the limit of very deep wells, when we have seen that the solid is then equivalent to N independent wells. Each band then consists of an infinitely narrow band of N levels of the same energy (recall Fig. 8.44), and this number is preserved when interactions are allowed between the wells. The conclusion we have just drawn concerning the number of levels in a band is of the greatest importance for understanding the electronic structure of solids, for it implies that each Brillouin zone can accommodate up to 2N electrons. When each atom provides one electron, the zone is only half full and the solid is a metallic conductor. When each atom provides two electrons, the lowest zone is full and there is an energy gap before the next zone becomes available; such a material is a semiconductor (and an effective insulator if the gap is large). This description mirrors exactly the conclusions of the molecular orbital, tight-binding description of solids.
PROBLEMS 8.1 The dependence of the molecular integrals j 0 , k 0 , and S for the hydrogen molecule–ion on the internuclear separation R are specified in Section 8.3. Plot the variation of the integrals a and b and the energies Eþ and E against R and identify the equilibrium bond length and the dissociation energy of the molecule–ion.
8.6 All the integrals involved in the H2 molecular orbital calculation are listed in eqn 8.29 and Further information 10. (a) Write and run a procedure using mathematical software to calculate E 2E1s as a function of R. (b) Identify the equilibrium bond length and the dissociation energy.
8.2 Confirm that 12(E þ Eþ) E1s is a positive quantity, and hence that the effect of an antibonding orbital outweighs the effect of a bonding orbital. Hint. Set up expressions for the quantity using eqns 8.23 and 8.24 and the results of Example 8.1; proceed to plot the quantity against R.
8.7 Evaluate the probability density for a single electron at a point on a line running between the two nuclei in H2 and plot the difference density r1 2(ca2 þ cb2) for R ¼ 74 pm. Hint. Use eqn 8.27. The probability density of electron 1, r1, is obtained from c2(1,2) by integrating over all locations of electron 2, because the latter’s Rposition is irrelevant. Therefore, begin by forming r1 ¼ c2(1, 2) dt2.
8.3 (a) Evaluate the probability density of the electron in Hþ 2 at the mid-point of the bond, and plot it as a function of R. (b) Evaluate the difference densities r ¼ c2 12 ðc2a þ c2b Þ at points along the line joining the two nuclei (including the regions outside the nuclei) for R ¼ 130 pm. The difference density shows the modification to the electron distribution brought about by constructive (or destructive) overlap. Hint. Use the c in eqn 8.15. The overlap integral S is given in Example 8.1. (c) Repeat the calculation for several values of R. 8.4 We shall see in Chapter 10 that the vibrational frequency of a chemical bond is o ¼ (k/m)1/2, where k ¼ (d2E/dR2)0 is the force constant and m is the effective mass; for a homonuclear diatomic molecule of atoms of mass m, m ¼ 12m. Estimate the vibrational frequency of the hydrogen molecule–ion. 8.5 Take the hydrogen molecule wavefunction in eqn 8.27 and find an expression for the expectation value of the hamiltonian in terms of molecular integrals. Hint. The outcome of this calculation is eqn 8.28.
8.8 Confirm that the CI wavefunction c ¼ c1C1 þ c3C3 in Section 8.5 can be expressed as shown there, in terms of the sums and differences of the coefficients ci. 8.9 Predict the ground configuration of (a) C2, (b) Cþ 2, (c) C2 , (d) N2þ , (e) N2 , (f) F2þ , and (g) Ne2þ . Decide which terms can arise in each case, and suggest which lies lowest in energy. 8.10 The bond order of a diatomic molecule can be determined from its molecular orbital energy level diagram by taking the difference between the numbers of bonding and antibonding electrons and dividing by two. Compute the bond orders of all the species in Problem 8.9. 8.11 Predict the ground configuration of (a) CO, (b) NO. Decide which terms can arise in each case, and suggest which lies lowest in energy. 8.12 Use a minimal basis set for the MO description of the molecule H2O to show that the secular determinant factorizes into (1 1), (2 2), and (3 3) determinants.
286
j
8 AN INTRODUCTION TO MOLECULAR STRUCTURE
Set up the secular determinant, denoting the Coulomb integrals aH, aO, and aO 0 for H1s, O2p, and O2s, respectively, and writing the (O2p, H1s) and (O2s, H1s) resonance integrals as b and b 0 , respectively. Neglect overlap. First, neglect the 2s-orbital, and find expressions for the energies of the molecular orbitals for a bond angle of 90 . 8.13 Now develop the previous calculation by taking into account the O2s-orbital. Set up the secular determinant with the bond angle y as a parameter. Find expressions for the energies of the molecular orbitals and of the entire molecule. As a first step in analysing the expressions, set aH aO
aO 0 and b b 0 . Can you devise improvements to the values of the Coulomb integrals on the basis of atomic spectral data? 8.14 Set up and solve the secular determinants for (a) hexatriene, (b) the cyclopentadienyl radical in the Hu¨ckel pelectron scheme; find the energy levels and molecular orbitals, and estimate the delocalization energy. 8.15 (a) Confirm that the symmetry-adapted linear combinations of p-orbitals for benzene are those set out above Example 8.4. (b) Find the corresponding combinations for naphthalene. 8.16 Within the Hu¨ckel p-electron scheme, estimate the delocalization energy of (a) the benzene cation C6 Hþ 6 and (b) the benzene dianion C6H26 . 8.17 The allyl radical CH2 ¼ CHCH2 is a conjugated p-system having a p-orbital on the carbon atom adjacent to a double bond. Estimate its p-electron energy by using the Hu¨ckel approximation. 8.18 Confirm that the solutions of a tridiagonal determinant are those given in Example 8.3. 8.19 Show that the roots of the secular determinant for a cyclic polyene of N atoms can be constructed by inscribing a regular N-gon in a circle and noting the locations of the corners of the polygon, as in Fig. 8.31. Hint. See A.A. Frost and B. Musulin. J. Chem. Phys., 21, 572 (1953).
and 1 the elements of the determinant become o ¼ (a E)/(b ES) and 1, respectively. Hence the roots in terms of o are the same as the roots in terms of x. Solve for E. Typically S ¼ 0.25. 8.22 Find the effect of including neighbouring atom overlap on the p-electron energy levels of benzene. Use a computer to explore how the energies depend on the bond lengths, using b / S and 1 3 Sð2pp, 2ppÞ ¼ 1 þ s þ 25 s2 þ 15 s es
s¼
Zeff R 2a0
where Zeff is taken from Table 7.1. Consider the difference in resonance energy between the cases where the molecule has six equivalent C–C bond lengths of 140 pm (the experimental value) and where it has alternating lengths of 133 pm and 153 pm (typical C¼ ¼C and C C lengths, respectively). 8.23 Determine which symmetry species are spanned by d-orbitals in a tetrahedral complex. 8.24 An ion with the configuration f 2 enters an environment of octahedral symmetry. What terms arise in the free ion, and with which terms do they correlate in the complex? Hint. Follow the discussion of Section 8.11. 8.25 In the strong field case, the configuration d2 gives 1 1 2 rise to e2g , t2g eg , and t2g . (a) What terms may arise? (b) How do the singlet terms of the complex correlate with the singlet terms of the free ion? (c) What configurations arise in a tetrahedral complex, and what are the correlations? 8.26 Find the symmetry-adapted linear combinations of (a) s-orbitals, (b) p-orbitals on the ligands of an octahedral complex. Hint. Set Cartesian axes on each ligand site, with z pointing towards the central ion, determine how the orbitals are transformed under the operations of the group O, and use the procedures for establishing symmetry-adapted orbitals as described in Chapter 5.
8.20 Heterocyclic molecules may be incorporated into the Hu¨ckel scheme by modifying the Coulomb integral of the atom concerned and the resonance integrals to which it contributes. Consider pyridine, C5H5N (symmetry group C2v). Construct and solve the Hu¨ckel secular determinant with bCC bCN b and aN ¼ (aC þ 12b). Estimate the electron energy and the delocalization energy. Hint. The roots of the determinants are best found on a computer.
8.27 Repeat Problem 8.26 for a tetrahedral complex. What is the role of p-bonding in such complexes?
8.21 Explore the role of p-orbital overlap in p-electron calculations. Take the cyclobutadiene secular determinant, but construct it without neglect of overlap between neighbouring atoms. Show that in place of x ¼ (a E)/b
8.30 Explore the effect of changing the depth of the potential well by finding the solutions of eqn 8.46 for different values of g. Solve the equations numerically for (a) g ¼ p and (b) g ¼ 2p.
8.28 Repeat Problem 8.26 for a square planar complex. What is the role of p-bonding in such complexes? 8.29 Verify that the Kronig–Penney model results in eqn 8.44, and show that this condition can be expressed as eqn 8.45.
9 The Hartree–Fock self-consistent field method 9.1 The formulation of the approach 9.2 The Hartree–Fock approach 9.3 Restricted and unrestricted Hartree–Fock calculations 9.4 The Roothaan equations 9.5 The selection of basis sets 9.6 Calculational accuracy and the basis set Electron correlation 9.7 Configuration state functions 9.8 Configuration interaction 9.9 CI calculations 9.10 Multiconfiguration and multireference methods 9.11 Møller–Plesset many-body perturbation theory 9.12 The coupled-cluster method Density functional theory 9.13 Kohn–Sham orbitals and equations 9.14 Exchange–correlation functionals Gradient methods and molecular properties 9.15 Energy derivatives and the Hessian matrix 9.16 Analytical derivatives and the coupled perturbed equations Semiempirical methods 9.17 Conjugated p-electron systems 9.18 Neglect of differential overlap Molecular mechanics 9.19 Force fields 9.20 Quantum mechanics–molecular mechanics Software packages for electronic structure calculations
The calculation of electronic structure
One of the primary goals in molecular quantum mechanics is to solve the time-independent Schro¨dinger equation and to determine the electronic structures of atoms and molecules. Chapter 8 established the qualitative features of molecular structure calculations in terms of visualizable concepts. In this chapter, we introduce some of the computational techniques that are used to solve the Schro¨dinger equation for electrons in molecules: we establish the equations that are used and describe some of the approximations that make the computations feasible. Our starting point is the Born–Oppenheimer approximation (Section 8.1) and our focus is the solution of the electronic Schro¨dinger equation Hcðr; RÞ ¼ EðRÞcðr; RÞ
ð9:1Þ
for a fixed set of locations R of the nuclei. The electronic wavefunction c depends on the electronic coordinates r and parametrically on R; E(R) is the electronic energy. The hamiltonian is H¼
n n X N n X X 2 X h ZI e2 e2 r2i þ 12 2me i 4pe0 rIi 4pe0 rij i ij I
ð9:2Þ
In molecular structure calculations it is conventional not to include the nucleus–nucleus repulsion term in H, but to add it as a classical term at the end of the calculation; we adopt that convention here. There are two main approaches to the solution of the Schro¨dinger equation. In an ab initio calculation,1 a model is chosen for the electronic wavefunction and eqn 9.1 is solved using as input only the values of the fundamental constants and the atomic numbers of the nuclei. The accuracy of this approach is determined primarily by the model chosen for the wavefunction. For large molecules, accurate ab initio calculations are computationally expensive and semiempirical methods have been developed to treat a wider variety of chemical species. A semiempirical method makes use of a simplified form for the hamiltonian and adjustable parameters obtained from experimental data. In both cases it is a challenging task to compute ‘chemically accurate’ energies; that is, energies calculated within about 0.05 eV (about 5 kJ mol1) of the exact values. This chapter concentrates on the calculation of the electronic wavefunction and the electronic energy. However, once those quantities are known, a wide .......................................................................................................
1. The term ab initio comes from the Latin words for ‘from the beginning’.
288
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
range of chemically and physically important properties can be determined. For example, by finding the minimum of the potential energy surface (that is, the electronic energy plus the nucleus–nucleus repulsion energy) of a stable molecule, it is possible to characterize its equilibrium structure in terms of its bond lengths and bond angles. Force constants and vibrational frequencies can be determined from gradients of the potential energy surfaces. The modern trend is to broaden the range of properties that are calculated to include the location of the stationary points of the surface describing a reactive system, and hence to characterize activated complexes and transition states. Almost universally in the field of electronic structure calculations, but not in this presentation, the equations and results are expressed in ‘atomic units’ (au). Atomic units do not have the same status as actual SI units, as they are combinations of various fundamental constants and have values that change when the accepted values of the fundamental constants change. In atomic units, lengths are expressed as multiples of the bohr (the Bohr radius) h2/mee2 ( ¼ 52.918 pm) and energies are expressed in terms of the a0 ¼ 4pe0 h2=me a20 ¼ 2hcR1 ð¼ 4:3597 aJ, 27:21 eVÞ: hartree, Eh ¼
The Hartree–Fock self-consistent field method The self-consistent field method was described in Section 7.15. However, because it is the starting point of many of the ab initio methods, we consider it again here in more detail and generalize some of the previous discussion.
9.1 The formulation of the approach The crucial complication in all electronic structure calculations is the presence of the electron–electron potential energy, which depends on the electron– electron separations rij as given by the third term in eqn 9.2. As a first step, suppose that the true electronic wavefunction, c, is similar in form to the wavefunction c that would be obtained if this complicating feature were neglected. That is, c is a solution of n X H ¼ hi ð9:3Þ H c ¼ E c i¼1
where hi is the core hamiltonian for electron i (see Section 7.15). This n-electron equation can be separated into n one-electron equations, so we can immediately write c as a product of n one-electron wavefunctions (orbitals) of the form ca ðr i ; RÞ. To simplify the notation, we shall denote the orbital occupied by electron i with coordinate ri and parametrically depending on the nuclear arrangement R as ca ðiÞ. It is a solution of hi ca ðiÞ ¼ Ea ca ðiÞ
ð9:4Þ
Ea
is the energy of an electron in orbital a in this independent-electron where model. The overall wavefunction c is a product of one-electron wavefunctions: c ¼ ca ð1Þcb ð2Þ . . . cz ðnÞ
ð9:5Þ
9.2 THE HARTREE–FOCK APPROACH
j
289
The function c depends on all the electron coordinates and, parametrically, on the nuclear locations. The overall energy E is a sum of the one-electron energies. At this stage, we have not taken into account the spin of the electron or the requirement that the electronic wavefunction must obey the Pauli principle. To do so, we introduce the concept of the spinorbital, fa(i), first encountered in Section 7.11. A spinorbital is a product of an orbital wavefunction and a spin function, and in a more elaborate notation would be denoted fa(xi;R), where xi represents the joint spin–space coordinates of electron i. So that the Paul principle is obeyed, we use a Slater determinant (Section 7.11) and the overall wavefunction is written as: c ðx; RÞ ¼ ðn!Þ1=2 detjfa ð1Þfb ð2Þ . . . fz ðnÞj
ð9:6Þ
The spinorbitals fu, with u ¼ a, b, . . . , z, are orthonormal and the label u now incorporates the spin state as well as the spatial state.
9.2 The Hartree–Fock approach Electron–electron repulsions are critically important and must be included in any accurate electronic structure treatment. In the Hartree–Fock method (HF method),2 a product wavefunction of the form of eqn 9.6 is sought, with the electron–electron repulsions treated in an ‘average’ way. Each electron is considered to be moving in the electrostatic field of the nuclei and the average field of the other n 1 electrons. The spinorbitals that give the ‘best’ n-electron determinantal wavefunction are found by using variation theory (Section 6.9), which involves minimizing the Rayleigh ratio R
c ðx; RÞHcðx; RÞ dx ð9:7Þ e¼ R
c ðx; RÞcðx; RÞ dx subject to the constraint that the spinorbitals are orthonormal. The lowest value of e is identified with the electronic energy of the ground state of the atom and molecule for the selected nuclear configuration R. The application of this minimization procedure leads to the Hartree– Fock equations for the individual spinorbitals (see Further information 11). The Hartree–Fock equation for spinorbital fa(1), where we have arbitrarily assigned electron 1 to spinorbital fa, is f1 fa ð1Þ ¼ ea fa ð1Þ where ea is the spinorbital energy and f1 is the Fock operator: X f Ju ð1Þ Ku ð1Þg f1 ¼ h1 þ
ð9:8Þ
ð9:9Þ
u
In eqn 9.9, h1 is the core hamiltonian for electron 1, the sum is over all spinorbitals u ¼ a, b, . . . , z, and the Coulomb operator, Ju, and exchange .......................................................................................................
2. Electronic structure calculations are littered with acronyms: the terms we use in this chapter are collected together in Box 9.1 at the end of the chapter.
290
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
operator, Ku, are defined as follows: Z 1 f u ð2Þ fu ð2Þ dx2 fa ð1Þ Ju ð1Þfa ð1Þ ¼ j0 r12 Z 1 f u ð2Þ Ku ð1Þfa ð1Þ ¼ j0 fa ð2Þ dx2 fu ð1Þ r12
ð9:10aÞ ð9:10bÞ
where, as in eqn 8.19, e2 j0 ¼ 4pe0 The Coulomb and exchange operators are defined here in terms of spinorbitals rather than in terms of spatial wavefunctions, as in Section 7.15,3 but their meaning is essentially the same: the Coulomb operator takes into account the Coulombic repulsion between electrons, and the exchange operator represents the modification of this energy that can be ascribed to the effects of spin correlation. It follows that the sum in eqn 9.9 represents the average potential energy of electron 1 due to the presence of the other n 1 electrons. Note that because Ja ð1Þfa ð1Þ ¼ Ka ð1Þfa ð1Þ the sum in eqn 9.9 includes contributions from all spinorbitals fu except the fa being computed. Each spinorbital must be obtained by solving an equation of the form of eqn 9.8 with the corresponding Fock operator fi. However, because fi depends on the spinorbitals of all the other n 1 electrons, it appears that to set up the HF equations, one must already know the solutions beforehand! This is a common dilemma in electronic structure calculations, and it is commonly attacked by adopting an iterative style of solution, and stopping when the solutions are self-consistent; hence the name self-consistent field (SCF) is given to this approach. In a self-consistent procedure, a trial set of spinorbitals is formulated and used to construct the Fock operator, then the HF equations are solved to obtain a new set of spinorbitals that are used to construct a revised Fock operator, and so on. The cycle of calculation and reformulation is repeated until a convergence criterion is satisfied.4 The Fock operator defined in eqn 9.9 depends on the n occupied spinorbitals. However, once these spinorbitals have been determined, the Fock operator can be treated as a well-defined hermitian operator and, like other hermitian operators (for example the hamiltonian operator), it has an infinite number of eigenfunctions. In other words, there is an infinite number of spinorbitals fu with an accompanying energy eu that solve eqn 9.8. In practice, of course, we have to be content with solving eqn 9.8 for a finite number m of spinorbitals with m n. .......................................................................................................
3. To compare eqn 9.9 with eqn 7.47, we need to note that the factor of 2 accompanying J in the latter equation arises when the Hartree–Fock equations for the spinorbitals are converted to equations for spatial orbitals, with each spatial orbital doubly occupied. 4. Convergence problems are sometimes encountered but they usually are not a major problem for many calculations. Several methods have been developed to improve convergence; these include the ‘level shifter method’ of V.R. Saunders and I.H. Hillier, Int. J. Quantum Chem., 699, 7 (1973), and the direct inversion of the iterative subspace, P. Pulay, Chem. Phys. Lett., 393, 73 (1980).
j
9.3 RESTRICTED AND UNRESTRICTED HARTREE–FOCK CALCULATIONS
291
The m optimized spinorbitals obtained on completion of the HF-SCF procedure are arranged in order of increasing orbital energy, and the n lowest energy spinorbitals are called the occupied orbitals. The remaining unoccupied m n spinorbitals are called virtual orbitals. The Slater determinant (of the form given in eqn 9.6) composed of the n occupied spinorbitals is the HF ground-state wavefunction for the molecule; we shall denote it F0.5 By ordering the orbital energies and analysing the radial and angular nodal patterns of the spatial parts of the spinorbitals, we can identify a spinorbital as a 1s-spinorbital, a 2s-spinorbital, and so on.
9.3 Restricted and unrestricted Hartree–Fock calculations It is customary in HF-SCF calculations on closed-shell states of atoms (for which the number of electrons, n, is always even) to suppose that the spatial components of the spinorbitals are identical for each member of a pair of electrons. There are then 12n spatial orbitals of the form ca(r1) and the HF wavefunction is F0 ¼ ðn!Þ1=2 det jcaa ð1Þcba ð2Þcab ð3Þ . . . cbz ðnÞj
ð9:11Þ
where we have used the same notation as in Section 7.11. Such a wavefunction is called a restricted Hartree–Fock (RHF) wavefunction. The HF equations for the spinorbitals given in eqns 9.8–9.10 are converted to the set of spatial eigenvalue equations given in eqns 7.47–7.49 by integration over the spin functions and using the orthonormality of a and b.6 Two procedures are commonly used for open-shell states of atoms. In the restricted open-shell formalism, all electrons except those occupying open-shell orbitals are forced to occupy doubly occupied spatial orbitals. For example, the restricted open-shell wavefunction for atomic lithium would be of the form F0 ¼ ð6Þ1=2 det jca1s ð1Þcb1s ð2Þca2s ð3Þj in which the first two spinorbitals in the Slater determinant (which we identify as 1s-spinorbitals) have the same spatial wavefunction. However, the restricted open-shell formalism imposes a severe constraint on the wavefunction; whereas the 1sa-electron has an exchange interaction with the 2sa-electron, the 1sb-electron does not and, as a result, the variational ground-state energy is usually not accurate. In the unrestricted open-shell Hartree–Fock (UHF) formalism the two 1s-electrons are not constrained to the same spatial wavefunction. For instance, the UHF wavefunction for Li would be of the form F0 ¼ ð6Þ1=2 det jcaa ð1Þcbb ð2Þcac ð3Þj .......................................................................................................
5. The HF ground-state wavefunction, F0, is either a single determinant or a linear combination of a small number of Slater determinants chosen to give the correct symmetry of the electronic state. 6. The details of the conversion are given in Section 3.4.1 of the excellent book A. Szabo and N.S. Ostlund, Modern quantum chemistry: Introduction to advanced electronic structure, Dover Publications, Inc., New York (1996). This text should be consulted for many of the details of the discussions in this chapter.
292
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
in which all three spatial orbitals are different (with ca and cb versions of 1s-orbitals and cc a 2s-orbital). By relaxing the constraint of occupying orbitals in pairs, the open-shell UHF formalism gives a lower variational energy than the open-shell RHF formalism. However, one disadvantage of the UHF approach is that whereas the RHF wavefunction is an eigenfunction of S2, the UHF function is not; that is, the total spin angular momentum is not well-defined for a UHF wavefunction. Example 9.1 Showing that the RHF wavefunction is an eigenfunction of S2
Consider the following restricted Hartree–Fock wavefunction for the helium atom: F0 ¼ ð2Þ1=2 detjcaa ð1Þcba ð2Þj Show that this Slater determinant is an eigenfunction of S2 and evaluate its eigenvalue. Method. We need to expand the Slater determinant and consider the effect
of the spin operator, which acts only on the spin states a and b and not on the spatial function ca. Because we are dealing with a two-electron system, S ¼ s1 þ s2, where si acts only on electron i. We use the relations S2 ¼ ðs1 þ s2 Þ ðs1 þ s2 Þ ¼ s21 þ s22 þ 2s1 s2 and s1 s2 ¼ s1z s2z þ s1x s2x þ s1y s2y ¼ s1z s2z þ 12 ðs1þ s2 þ s1 s2þ Þ The results of the operations of s2, sz, sþ , and s on a and b are given in Section 4.8. Answer. First, we expand the determinant:
F0 ¼ ð2Þ1=2 fca ð1Það1Þca ð2Þbð2Þ ca ð2Það2Þca ð1Þbð1Þg The effect of S2 on the first term in F0 is S2 ca ð1Það1Þca ð2Þbð2Þ ¼ ca ð1Þs21 að1Þca ð2Þbð2Þ þ ca ð1Það1Þca ð2Þs22 bð2Þ þ 2ca ð1Þs1z að1Þca ð2Þs2z bð2Þ þ ca ð1Þs1þ að1Þca ð2Þs2 bð2Þ þ ca ð1Þs1 að1Þca ð2Þs2þ bð2Þ ¼ 34 h2 ca ð1Það1Þca ð2Þbð2Þ þ 34 h2 ca ð1Það1Þca ð2Þbð2Þ 12 h2 ca ð1Það1Þca ð2Þbð2Þ þ 0 þ h2 ca ð1Þbð1Þca ð2Það2Þ ¼ h2 ca ð1Það1Þca ð2Þbð2Þ þ h2 ca ð1Þbð1Þca ð2Það2Þ A similar analysis of the effect of S2 on the second term in F0 yields S2 ca ð2Það2Þca ð1Þbð1Þ ¼ h2 ca ð2Það2Þca ð1Þbð1Þ þ h2 ca ð2Þbð2Þca ð1Það1Þ On collecting terms we obtain n S2 F0 ¼ ð2Þ1=2 ðh2 h2 Þca ð1Það1Þca ð2Þbð2Þ
o h2 Þca ð2Það2Þca ð1Þbð1Þ ¼ 0 ðh2
9.4 THE ROOTHAAN EQUATIONS
j
293
Therefore, F0 is an eigenfunction of S2 with an eigenvalue of zero, as is to be expected because the ground state of the closed-shell helium atom is a singlet. Self-test 9.1. Confirm that the UHF wavefunction for helium of the form F0 ¼
(2)1=2 detjca(1)a(1)cb(2)b(2)j, where ca 6¼ cb, is not an eigenfunction of S2.
In practice, the expectation value of S2 for the unrestricted wavefunction is computed and compared with the true value S(S þ 1) h2 for the ground state. If the discrepancy is not significant, the UHF method has given a reasonable molecular wavefunction. The UHF wavefunction is often used as a first approximation to the true wavefunction even if the discrepancy is significant. It is also possible to use projection operator techniques on the UHF wavefunction to obtain an improved wavefunction with more accurate S2 expectation values.
9.4 The Roothaan equations We have concealed a difficulty up to this point. The HF-SCF procedure is relatively straightforward to implement for atoms, for their spherical symmetry means that the HF equations can be solved numerically for the spinorbitals. However, such numerical solution for spinorbitals for molecular systems is sufficiently computationally complex that a modification of the technique must be used. As long ago as 1951, C.C.J. Roothaan and G.G. Hall independently suggested using a known set of basis functions with which to expand the spinorbitals (more precisely, the spatial parts of the spinorbitals). In this section, which is limited to a discussion of the restricted closedshell Hartree–Fock formalism, we show how this suggestion transforms the coupled HF equations into a matrix problem which can be solved by using matrix manipulations. We begin with eqn 7.47 for the spatial function ca(1) occupied by electron 1 and write it in the notation of this chapter as f1 ca ð1Þ ¼ ea ca ð1Þ
ð9:12Þ
where f1 is the Fock operator expressed in terms of the spatial wavefunctions: X ð9:13Þ f1 ¼ h1 þ f2Ju ð1Þ Ku ð1Þg u
and the Coulomb and exchange operators are defined in eqns 7.48 and 7.49, solely in terms of spatial coordinates. Next, we introduce a set of M basis functions yj (their form is described in Section 9.5) and express each spatial wavefunction ci as a linear combination of these basis functions: ci ¼
M X
cji yj
ð9:14Þ
j¼1
where cji are as yet unknown coefficients. From a set of M basis functions, we can obtain M linearly independent spatial wavefunctions, and the problem
294
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
of calculating the wavefunctions has been transformed to one of computing the coefficients cji. When the expansion in eqn 9.14 is substituted into eqn 9.12, we obtain f1
M X
cja yj ð1Þ ¼ ea
j¼1
M X
cja yj ð1Þ
ð9:15Þ
j¼1
Multiplication of both sides of this equation by y i ð1Þ and integration over r1 yields M X
cja
Z
y i ð1Þf1 yj ð1Þ dr 1 ¼ ea
j¼1
M X
cja
Z
y i ð1Þyj ð1Þ dr 1
ð9:16Þ
j¼1
As is often the case in quantum chemistry, the structure of a set of equations becomes clearer if we introduce a more compact notation. In this case, it proves sensible to introduce the overlap matrix, S, with elements Z ð9:17Þ Sij ¼ y i ð1Þyj ð1Þ dr 1 (this matrix is not in general the unit matrix because the basis functions are not necessarily orthogonal) and the Fock matrix, F, with elements Z Fij ¼ y i ð1Þf1 yj ð1Þ dr 1 ð9:18Þ Then eqn 9.16 becomes M X
Fij cja ¼ ea
j¼1
M X
Sij cja
ð9:19Þ
j¼1
This expression is one in a set of M simultaneous equations (one for each value of i) that are known as the Roothaan equations. The entire set of equations can be written as the single matrix equation Fc ¼ Sc"
ð9:20Þ
where c is an M M matrix composed of elements cja and " is an M M diagonal matrix of the orbital energies ea. At this stage we can make progress by drawing on some of the properties of matrix equations (see Further information 23). The Roothaan equations have a non-trivial solution only if the following secular equation is satisfied: detjF ea Sj ¼ 0
ð9:21Þ
This equation cannot be solved directly because the matrix elements Fij involve integrals over the Coulomb and exchange operators which themselves depend on the spatial wavefunctions. Therefore, as before, we must adopt a self-consistent field approach, obtaining with each iteration a new set of coefficients cja and continuing until a convergence criterion has been reached (Fig. 9.1). It is instructive to examine the matrix elements of the Fock operator, for in that way we can begin to appreciate some of the computational difficulties of
9.4 THE ROOTHAAN EQUATIONS
Choose set of basis functions j
j
295
Formulate set of trial coefficients cja (and therefore wavefunctions a
eqn 9.17
eqn 9.18
Fock matrix, F
Overlap matrix, S
Done No
Yes
eqn 9.21 Convergence?
Fig. 9.1 A summary of the iteration procedure for a Hartree–Fock self-consistent field calculation.
Energies, a coefficients, cja
obtaining HF-SCF wavefunctions. The explicit form of the matrix element Fij is obtained from eqns 9.13, 7.48, and 7.49, and is Fij ¼
Z
y i ð1Þh1 yj ð1Þ dr 1
þ 2j0
XZ
y i ð1Þc u ð2Þ
u
j0
XZ
y i ð1Þc u ð2Þ
u
1 c ð2Þyj ð1Þ dr 1 dr 2 r12 u
ð9:22Þ
1 yj ð2Þcu ð1Þ dr 1 dr 2 r12
The first term on the right is a one-electron integral that we shall denote hij. Insertion of the expansion in eqn 9.14 results in the following expression for Fij solely in terms of integrals over the known basis functions: X
Fij ¼ hij þ 2j0
c lu cmu
Z
y i ð1Þy l ð2Þ
u;l;m
j0
X
c lu cmu
u;l;m
Z
y i ð1Þy l ð2Þ
1 ym ð2Þyj ð1Þ dr 1 dr 2 r12
1 yj ð2Þym ð1Þ dr 1 dr 2 r12
ð9:23Þ
The appearance of this rather horrendous expression can be greatly simplified by introducing the following notation for the two-electron integrals over the basis functions: Z 1
ðabjcdÞ ¼ j0 y a ð1Þyb ð1Þ y ð2Þyd ð2Þ dr 1 dr 2 ð9:24Þ r12 c Equation 9.23 then becomes X Fij ¼ hij þ c lu cmu f2ðijjlmÞ ðimjljÞg u;l;m
ð9:25Þ
296
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
which is usually written as X Fij ¼ hij þ Plm ðijjlmÞ 12ðimjljÞ
ð9:26Þ
l;m
where Plm is defined as X Plm ¼ 2 c lu cmu
ð9:27Þ
u
The matrix elements Plm are referred to as density matrix elements, and are interpreted as the total electron density in the overlap region of yl and ym; recall that the summation in eqn 9.27 is over all spatial wavefunctions. When l ¼ m, Pll is the electron density on atom l; when l 6¼ m, Plm is the bond order between l and m. The one-electron matrix elements hij need to be evaluated only once because they remain unchanged during each iteration. However Plm, which depends on the expansion coefficients clu and cmu, does need to be re-evaluated at each iteration. Because the number of two-electron integrals (eqn 9.24) to evaluate is of the order of M4—so even small basis sets for moderately-sized molecules can rapidly approach millions of two-electron integrals—their efficient calculation poses the greatest challenge in an HF-SCF calculation. The problem is alleviated somewhat by the possibilities that a number of integrals may be identically zero due to symmetry, some of the non-zero integrals may be equal by symmetry, and some of the integrals may be negligibly small because the basis functions may be centred on atomic nuclei separated by a large distance. Nevertheless, in general, there will be many more two-electron integrals than can be stored in the core memory of the computer, and a large body of work has been done in trying to develop efficient approaches to the calculation of two-electron integrals.7
9.5 The selection of basis sets In principle, a complete set of basis functions must be used to represent spinorbitals exactly, and the use of an infinite number of basis functions would then result in a Hartree–Fock energy equal to that given by the variational expression, eqn 9.7. This limiting energy is called the Hartree– Fock limit. The HF limit is not the exact ground-state energy of the molecule because it still ignores effects of electron correlation (a point discussed below). However, because an infinite basis set is not computationally feasible, a finite basis set is always used and the error due to the incompleteness of the basis is called the basis-set truncation error. A measure of this error is the difference between the Hartree–Fock limit and the computed lowest energy in an HF-SCF calculation. A critical computational consideration therefore will be to keep the number of basis functions low (to minimize the number of .......................................................................................................
7. For a discussion of some of these approaches, see Section 2.3 of D.M. Hirst, A computational approach to chemistry, Blackwell Scientific Publications, Oxford (1990) and references therein. Another useful reference is C.M. Quinn, Computational quantum chemistry: an interactive guide to basis set theory, Academic Press, London (2002).
9.5 THE SELECTION OF BASIS SETS
j
297
000
200
r
0 100
(a)
(b)
(c)
(d)
Fig. 9.2 Gaussian orbitals. (a), (b), and (c) show the contour plots for s-, p-, and d-type Gaussians, respectively, with the 2 2 2 form er , xer , and xyer . (d) Cross-sections through the three wavefunctions.
two-electron integrals to evaluate) and to choose them cleverly (to minimize the computational effort for the evaluation of each integral), but nevertheless achieve a small basis-set truncation error. The basis functions chosen are usually real. One choice of basis functions for use in eqn 9.14 are the Slater-type orbitals (STOs) introduced in Section 7.14. A complete basis set consists of STOs with all permitted integral values of n, l, and ml and all positive values of the orbital exponents, z (zeta), the parameter that occurs in the radial part (ezr) of the STO. In practice, only a small number of all possible functions are used. The best values of z are determined by fitting STOs to the numerically computed atomic wavefunctions. For atomic SCF calculations, STO basis functions are centred on the atomic nucleus. For diatomic and polyatomic species, STOs are centred on each of the atoms. However, for HF-SCF calculations on molecules with three or more atoms, the evaluation of the many two-electron integrals (abjcd) is impractical. Indeed, this ‘two-electron integral problem’ was once considered to be one of the greatest problems in quantum chemistry. The introduction of Gaussian-type orbitals (GTOs) by S.F. Boys8 in 1950 played a major role in making ab initio calculations computationally feasible.9 Cartesian Gaussians are functions of the form 2
yijk ðr 1 r c Þ ¼ ðx1 xc Þi ðy1 yc Þj ðz1 zc Þk eajr 1 r c j
ð9:28Þ
where (xc, yc, zc) are the Cartesian coordinates of the centre of the Gaussian function at rc; (x1, y1, z1) are the Cartesian coordinates of an electron at r1; i, j, and k are non-negative integers; and a is a positive exponent. When i ¼ j ¼ k ¼ 0, the Cartesian Gaussian is an s-type Gaussian; when i þ j þ k ¼ 1, it is a p-type Gaussian; when i þ j þ k ¼ 2, it is a d-type Gaussian, and so on (Fig. 9.2). There are six d-type Gaussians. If preferred, six linear .......................................................................................................
8. S.F. Boys, Proc. R. Soc. (London), 542, A200 (1950). 9. A valuable review of the development and use of Gaussian basis sets in a variety of electronic structure techniques is A.F. Jalbout, F. Nazari, and L. Tucker, Theochem., 671, 1 (2004).
298
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
G1G 2
Amplitude
G1
G2
r
Fig. 9.3 The product of two Gaussians (G) is itself a Gaussian lying between the two original functions. In this illustration, the amplitude of the product has been multiplied by 100.
1
i
with the contraction coefficients dji and the parameters characterizing g held fixed during the calculation. The spatial orbitals are then expressed as a linear combination of the contracted Gaussians: X cji wj ð9:30Þ ci ¼
0.8 0.6
j
0.4
0.2
0
combinations of these d-type Gaussians can be used instead, five of them having the angular behaviour of the five real 3d-hydrogenic orbitals and the sixth being spherically symmetrical like an s-function. This sixth linear combination is sometimes eliminated from the basis set, but its elimination is not essential because we have not assumed that the basis set is orthogonal. Spherical Gaussians, in which factors like x1 xc are replaced by spherical harmonics, are also used. The central advantage of GTOs is that the product of two Gaussians at different centres is equivalent to a single Gaussian function centred at a point between the two centres (Fig. 9.3). Therefore, two-electron integrals on three and four different atomic centres can be reduced to integrals over two different centres, which are much easier to compute. However, there is also a disadvantage to using GTOs that to some extent negates the computational advantage. A 1s hydrogenic atomic orbital has a cusp at the atomic nucleus; an n ¼ 1 STO also has a cusp there, but a GTO does not (Fig. 9.4). Because a GTO gives a poorer representation of the orbitals at the atomic nuclei, a larger basis must be used to achieve an accuracy comparable to that obtained from STOs. To alleviate the latter problem, several GTOs are often grouped together to form what are known as contracted Gaussian functions. In particular, each contracted Gaussian, w, is taken to be a fixed linear combination of the original or primitive Gaussian functions, g, centred on the same atomic nucleus: X dji gi ð9:29Þ wj ¼
s-type Gaussian n = 1 STO
Distance from nucleus
Fig. 9.4 A hydrogenic 1s-orbital is an exponential function, so there is a cusp at the nucleus. A Gaussian does not have a cusp at the nucleus.
The use of contracted rather than primitive Gaussians reduces the number of unknown coefficients cji to be determined in the HF calculation. For example, if each contracted Gaussian is composed of three primitives from a set of 30 primitive basis functions, then whereas the expansion in eqn 9.14 involves 30 unknown cji coefficients, the corresponding expansion in eqn 9.30 has only 10 unknown coefficients. This decrease in the number of coefficients leads to potentially large savings in computer time with little loss of accuracy if the contracted Gaussians are well-chosen. How are the primitives and the contracted Gaussians constructed? In some applications, a set of basis functions is chosen and an atomic SCF calculation is performed, resulting in an optimized set of exponents (for example, a in eqn 9.28) for the basis functions, which can then be used in molecular structure calculations. The simplest type of basis set is a minimal basis set in which one function is used to represent each of the orbitals of elementary valence theory. A minimal basis set would include one function each for H and He (for the 1s-orbital); five basis functions each for Li to Ne (for the 1s-, 2s-, and three 2p-orbitals); nine functions each for Na to Ar, and so on. For instance, a minimal basis set for H2O consists of seven functions, and includes
9.5 THE SELECTION OF BASIS SETS
j
299
two basis functions to represent the two H1s orbitals, and one basis function each for the 1s-, 2s-, 2px-, 2py-, and 2pz-orbitals of oxygen. However, a minimal basis set results in wavefunctions and energies that are not very close to the Hartree–Fock limits: accurate calculations need more extensive basis sets. A significant improvement is achieved by adopting a double-zeta basis set (DZ basis set), in which each basis function in the minimal basis set is replaced by two basis functions. Compared to a minimal basis set, the number of basis functions has doubled and with it the number of variationally determined expansion coefficients cji. A DZ basis set for H2O, for instance, would use 14 functions. In a triple-zeta basis set (TZ basis set), three basis functions are used to represent each of the orbitals encountered in elementary valence theory. A split-valence basis set (SV basis set) is a compromise between the inadequacy of a minimal basis set and the computational demands of DZ and TZ basis sets. Each valence atomic orbital is represented by two basis functions while each inner-shell atomic orbital is represented by a single basis function. For example, for an atomic SCF calculation on C using contracted Gaussians in an SV basis, there is one contracted function representing the 1s-orbital, two representing the 2s-orbital, and two each for the three 2p-orbitals. The basis sets we have described so far ignore possible contributions from basis functions representing orbitals for which the value of the quantum number l is larger than the maximum value considered in elementary valence theory (such as the inclusion of d-orbitals in the discussion of carbon compounds). When bonds form in molecules, atomic orbitals are distorted (or polarized) by adjacent atoms. This distortion can be taken into account by including basis functions representing orbitals with high values of l. For example, the inclusion of p-type basis functions can model reasonably well the distortion of a 1s-orbital, and d-type functions are used to describe distortion of p-orbitals. The addition of these polarization functions to a DZ basis set results in a double-zeta plus polarization basis (DZP basis). For example, in a DZP basis for methane, a set of three 2p-functions is added to each hydrogen atom and a set of six 3d-functions is added to the carbon atom. There are numerous ways to construct contracted Gaussian basis sets. One approach is to make a least-squares fit of N primitive Gaussians to a set of STOs that have been optimized in an atomic SCF calculation. For example, an SCF calculation is performed on atomic carbon using STOs to find the contracted Gaussians best representing the 1s, 2s, and 2p STOs, and then these contracted Gaussians are used in a subsequent SCF calculation on methane. The expansion of an STO in terms of N primitive Gaussians is designated STO-NG. A common choice is N ¼ 3, giving a set of contracted Gaussians referred to as STO-3G. A second approach is to perform an atomic SCF calculation using a relatively large basis of Gaussian primitives. This procedure results in a set of variationally determined SCF coefficients (cji in eqn 9.14) for the primitives of each spatial orbital ci. The coefficients of the primitive Gaussians can then be used (eqn 9.29) to obtain contracted Gaussian basis sets for use in molecular calculations. In the (4s)/[2s] contraction scheme,10 .......................................................................................................
10. S. Huzinaga, J. Chem. Phys., 1293, 42 (1965).
300
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
four primitive s-type Gaussians are used to construct two basis set functions for atomic hydrogen. As in many contraction schemes, the most diffuse primitive (the one with the smallest value of the exponent a) is left uncontracted, and each of the remaining primitives appears in only one contracted Gaussian. That is, in the (4s)/[2s] scheme, three of the primitives are used to form one contracted Gaussian basis set function. In the (9s5p)/[3s2p] contraction scheme,11 nine s-type and five p-type primitive Gaussians (which have been optimized in an atomic SCF calculation on a Period 2 element) are contracted into three and two basis functions, respectively. This contraction scheme usually results in a split-valence basis set containing one basis function representing the inner-shell 1s-orbital, two basis functions for the valence 2s-orbital, and two for each of the three 2p-orbitals. Therefore, the contraction scheme reduces the total number of basis functions from 24 (five p-type primitives for each of 2px, 2py, and 2pz, and nine s-type primitives) to nine. This reduction achieves substantial decrease in computer time because the number of two-electron integrals to be evaluated is proportional to the fourth power of the number of basis functions. Other contraction schemes also result in valuable savings. In the 3-21G basis set,12 one contracted Gaussian composed of three primitives is used to represent each inner-shell atomic orbital. Each valence-shell orbital is represented by two functions, one a contracted Gaussian of two primitives and one a single (and usually diffuse) primitive. The primitives are first optimized in an SCF calculation on atoms, and the contracted sets are then used in molecular calculations. The 6-31G* basis set starts with the split-valence 6-31G basis and adds polarization functions in the form of six d-type functions for each atom other than H. Another star, an additional polarization function: 6-31G** indicates the addition to 6-31G of a set of three p-type polarization functions for each H atom. Example 9.2 Determining the number of basis set functions in a molecular structure calculation
Determine the total number of Gaussian basis set functions in an ab initio calculation of C2H2 using a 6-31G basis set. Method. Start with the 6-31G basis, which, for each atom in the molecule,
consists of (a) one contracted Gaussian composed of six primitives for each inner-shell orbital and (b) two functions for each valence-shell orbital, a contracted Gaussian of three primitives and a single uncontracted primitive. Then add six d-type polarization functions for each atom other than hydrogen. Answer. Each 1s-orbital of H is represented by two basis set functions, using a total of four primitive Gaussians. Each 1s-orbital of C is represented by one contracted Gaussian of six primitives. The 2s-, 2px-, 2py-, and 2pz-orbitals of
.......................................................................................................
11. T.H. Dunning, J. Chem. Phys., 2823, 53 (1970). 12. W.J. Hehre, L. Radom, P.v.R Schleyer, and J.A. Pople, Ab initio molecular orbital theory, Wiley, New York (1986).
9.6 CALCULATIONAL ACCURACY AND THE BASIS SET
j
301
each C atom are each represented by two basis set functions, one a contraction of three primitives and one a single uncontracted primitive. In addition, each C atom will also have six d-type polarization functions. Therefore, the total number of 6-31G basis set functions for C2 H2 is 2(1 þ 4 2 þ 6) þ 2 2 ¼ 34. The total number of primitives used is 2{6 þ 4 (3 þ 1) þ 6} þ 2 4 ¼ 64. Self-test 9.2. How many basis functions would there be in a molecular
structure calculation on H2O using a 6-31G
basis? [25 basis functions composed of 42 primitives]
The basis-set superposition error is a contribution to the inaccuracy of calculations that stems from the use of a finite basis set; this error may arise in the calculation of the interaction energy of two weakly bound systems. As an example, suppose we were interested in the energetics of the dimerization of hydrogen fluoride and we defined the interaction energy as the energy of the dimer minus the energies of the two infinitely separated monomers. If we used, for example, a 6-31G basis set for each of the atoms in hydrogen fluoride, it might seem that the obvious choice would be to use a 6-31G basis set on each of the four atoms of the dimer. However, when the energy of an individual hydrogen fluoride molecule is computed, only the basis set functions on two atoms (H and F) are used to describe each electronic spatial orbital c. On the other hand, the orbitals in the calculation for the dimer are expressed as linear combinations of the basis set functions on all four atoms. In other words, the basis set for the dimer is larger than that for either monomer, and this enlargement of the basis results in a non-physical lowering of the energy of the dissociated dimer relative to the separated monomers. A common method used to correct the basis-set superposition error is the counterpoise correction,13 in which the energies of the monomer are computed by using the full basis set used for the dimer. For example, in the case of the hydrogen fluoride dimer, when computing the energy of an individual molecule, a basis set is used that consists of functions centred on each nucleus of the monomer as well as the same basis set functions centred at the two points in space that would correspond to the equilibrium positions of the other two nuclei in the dimer.
9.6 Calculational accuracy and the basis set Table 9.1 presents results of ab initio Hartree–Fock SCF calculations on the ground states of several closed-shell molecules and shows how the SCF energy varies with the basis set used in the calculation. The reported SCF energies correspond to geometries at or nearly at equilibrium. The energies represent the sum of the electronic (ab initio) energy and the nucleus–nucleus repulsion energy for the selected geometry. We see from the table that the energy approaches the Hartree–Fock limit as the basis set becomes more complete. .......................................................................................................
13. S.F. Boys and F. Bernardi, Mol. Phys., 553, 19 (1970).
302
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
Table 9.1 Self-consistent field energies with a variety of basis sets
Basis set
H2
N2
CH4
NH3
H2O
STO-3G
1.117
107.496
39.727
55.454
74.963
4-31G
1.127
108.754
40.140
56.102
75.907
6-31G
1.127
108.942
40.195
56.184
76.011
6-31G
1.131
108.942
40.202
56.195
76.023
HF limit
1.134
108.997
40.225
56.225
76.065
The energies are expressed as multiples of the hartree, Eh ¼ 4.359 74 aJ.
Table 9.2 Self-consistent field equilibrium bond lengths with a variety of
basis sets
Basis set
H2
N2
CH4
NH3
H2O
STO-3G
1.346
2.143
2.047
1.952
1.871
4-31G
1.380
2.050
2.043
1.873
1.797
6-31G
1.380
2.039
2.048
1.897
1.791
6-31G
Observed
1.385
2.039
2.048
1.897
1.782
1.401
2.074
2.050
1.912
1.809
The bond lengths are expressed as multiples of the Bohr radius (a0 ¼ 52.917 72 pm).
We have mentioned that the electronic potential energy surface (or curve for a diatomic molecule) can be used to predict the equilibrium geometries of molecules. This prediction can then be compared directly with the best experimental values. Table 9.2 shows a number of calculated equilibrium bond lengths using the basis sets indicated in Table 9.1. A good ab initio SCF calculation typically is in error by 0.02–0.04a0 (corresponding to 1–2 pm).
Electron correlation However good the HF ground-state wavefunction F0 may appear to be, it is not the ‘exact’ wavefunction. The Hartree–Fock method relies on averages: it does not consider the instantaneous coulombic interactions between electrons; nor does it take into account the quantum mechanical effects on electron distributions because the effect of the n – 1 electrons on the electron of interest is treated in an average way. We summarize these deficiencies by saying that the HF method ignores electron correlation. A great deal of modern work in the field of electronic structure calculation is aimed at taking electron correlation into account.
9.8 CONFIGURATION INTERACTION
j
303
9.7 Configuration state functions The HF method yields a finite set of spinorbitals when a finite basis set expansion is used. In general, a basis with M members (see eqn 9.14) results in M spatial wavefunctions and 2M different spinorbitals. As discussed in Section 9.2, by ordering the spinorbitals energetically and taking the n lowest in energy (to be occupied by the n electrons), we form the Hartree–Fock wavefunction F0. However, there remain 2M n virtual orbitals. Clearly, many Slater determinants can be formed from the 2M spinorbitals; F0 is just one of them. By using the single determinantal wavefunction F0 as a convenient reference, we can classify all other determinants according to how many electrons have been promoted from occupied orbitals to virtual orbitals. To simplify the appearance of the following expressions, we write the normalized Slater determinants using the notation k . . . k, and hence denote F0 as
2n q
F0 ¼ ð1=n!Þ1=2 detjf1 f2 . . . fa fb . . . fn j ¼ jjf1 f2 . . . fa fb . . . fn jj
p
where fa and fb are among the n occupied spinorbitals for the Hartree–Fock ground state. A singly excited determinant corresponds to one for which a single electron in occupied spinorbital fa has been promoted to a virtual spinorbital fp (Fig. 9.5): Fap ¼ jjf1 f2 . . . fp fb . . . fn jj
c
A doubly excited determinant is one in which two electrons have been promoted, one from fa to fp and one from fb to fq
b
pq Fab ¼ jjf1 f2 . . . fp fq . . . fn jj
a
2 1 F0
p
Fa
pq
F ab
Fig. 9.5 The notation for excited determinants.
In a similar manner, we can form other multiply excited determinants. Each of the determinants, or a linear combination of a small number of them constructed so as to have the correct electronic symmetry,14 is called a configuration state function (CSF). More precisely, a CSF is an eigenfunction of all the operators that commute with H. These excited CSFs can be used to approximate excited-state wavefunctions or, as we shall now see, they can be used in a linear combination with F0 to improve the representation of the ground-state or any excited-state wavefunction.
9.8 Configuration interaction The exact ground-state or excited-state wavefunction can be expressed as a linear combination of all possible n-electron Slater determinants arising from a complete set of spinorbitals.15 Therefore, we can write the exact electronic wavefunction C for any state of the system in the form X X pq pq X pqr pqr C ¼ C0 F0 þ Cpa Fap þ Cab Fab þ Cabc Fabc þ ð9:31Þ a
14. For example, a linear combination of determinants for an eigenfunction of S . 15. A proof of this statement can be found in P.-O. Lo¨wdin, Adv. Chem. Phys., 207, 2 (1959), a classic review on electron correlation.
Number of configuration state functions
304
j
104
9 THE CALCULATION OF ELECTRONIC STRUCTURE
Full CI
Exact
103
102
10 Hartree-- Fock limit 1
5 10 15 20 25 30 Number of basis functions
Fig. 9.6 The approach to the exact energy of a system as the numbers of basis functions and configuration state functions increase.
where the Cs are expansion coefficients and where the limits in the summation indices (a < b and so on) ensure that we sum over all unique pairs of spinorbitals in doubly excited determinants, over all unique triplets of spinorbitals in triply excited determinants, and so on. In other words, a given excited determinant appears only once in the summation. An ab initio method in which the wavefunction is expressed in the form of eqn 9.31 is called configuration interaction (CI). A primitive example of CI was described in Section 8.5 in connection with the electronic structure of H2. The energy associated with the exact ground-state wavefunction of the form of eqn 9.31 is the exact non-relativistic ground-state energy (within the Born–Oppenheimer approximation). The difference between this exact energy and the HF limit is called the correlation energy. Configuration interaction accounts for the electron correlation neglected in the Hartree– Fock method. At this point the familiar refrain is inevitable: in practice, it is computationally impossible to handle an infinite basis set of n-electron Slater determinants with each determinant constructed from an infinite set of spinorbitals. Furthermore, it becomes computationally very demanding (both in computer time and storage) to handle extremely large numbers of determinants. The latter problem is slightly alleviated by the observation that a number of the determinants in eqn 9.31 can often be eliminated on the basis of symmetry. For example, if we are interested in computing an accurate wavefunction for the 1Sþ g ground state of H2, we do not need to include CSFs that do not correspond to the required 1Sþ g symmetry. For instance, we can ignore CSFs that have u parity or which have non-zero eigenvalues of Sz. There is another point that emphasizes—if further emphasis is required— the difficulty of carrying out molecular structure calculations reliably. Even if we could include all CSFs of the desired symmetry in eqn 9.31, we must also remember that the CSFs themselves are constructed from a finite set of oneelectron spinorbitals. A calculation is classified as full CI if all CSFs of the appropriate symmetry are used for a given finite basis set. For a given basis, full CI is the best CI calculation we can do. The difference between the ground-state energies obtained from a Hartree–Fock SCF calculation and a full CI calculation using the same basis set is called the basis-set correlation energy (Fig. 9.6). As the number of one-electron spinorbitals computed from the HF equations gets larger and larger, the basis-set correlation energy gets closer and closer to the exact correlation energy. Unfortunately, even for molecular calculations involving a small number (n) of electrons and a relatively small number (M) of basis set functions y (and consequently a small number 2M of spinorbitals), the total number of determinants can be extremely large. For example, with 10 electrons and 20 basis set functions, the total number of determinants is
2M ¼ 8:477 . . . 108 n In practice, therefore, the expansion in eqn 9.31 must almost always be truncated. Nonetheless, although the calculation is limited to a finite set of spinorbitals and a fraction of all possible determinants, CI is a popular
9.9 CI CALCULATIONS
j
305
method for the calculation of accurate molecular wavefunctions and potential energy surfaces. Even with a small number of CSFs it can correct for one of the deficiencies that stem from the use of only doubly occupied orbitals in the restricted HF method, namely the incorrect behaviour for the dissociation of a molecule. This point was illustrated in Section 8.5. A calculation in which the incorrect behaviour of the HF wavefunction upon dissociation is corrected accounts for an important part of the correlation energy called non-dynamical correlation or structural correlation. On the other hand, dynamical correlation accounts for the incorrect HF wavefunction at short interatomic distances.
9.9 CI calculations In configuration interaction calculations, the ground- or excited-state wavefunction, C, for state s (which we will denote for the remainder of this chapter by Cs to avoid confusion with the spatial function ci) is represented as a linear combination of n-electron Slater determinants. Equation 9. 31 can be written in a notationally simpler form as Cs ¼
L X
CJs FJ
ð9:32Þ
J¼1
where the sum is over a finite number L of determinants FJ with expansion coefficients CJs for the state s. The expansion coefficients CJs are determined variationally by minimizing the Rayleigh ratio e of eqn 9.7 using Cs as the trial function. As in all applications of variation theory (see Sections 6.9 and 6.10), this minimization is equivalent to solving a set of simultaneous equations for the coefficients CJs for each state s: L X
HIJ CJs ¼ Es
J¼1
L X
SIJ CJs
ð9:33Þ
J¼1
where HIJ ¼
Z
F I HFJ dx1 dx2 . . . dxn
ð9:34Þ
and SIJ ¼
Z
F I FJ dx1 dx2 . . . dxn
ð9:35Þ
R where the notation . . . dx implies integration over spatial (r) and spin coordinates. The set of equations can be written in matrix notation as HC ¼ ESC
ð9:36Þ
where the elements of the L L square matrices H and S are HIJ and SIJ, respectively; E is the diagonal matrix of energies Es; and C is an L L matrix of coefficients. Because the Slater determinants form an orthonormal set (SIJ ¼ dIJ), eqn 9.36 becomes HC ¼ EC
ð9:37Þ
306
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
Slater determinants, ΦI
Section 9.7
Spinorbitals, i
Section 7.11
Spatial wavefunctions, i
Section 9.4
Basis functions, i
Fig. 9.7 Matrix elements between Slater determinants can ultimately be0expressed in terms of matrix elements between basis functions.
This matrix equation can be solved by diagonalizing H, and yields a total of L wavefunctions (eigenfunctions) Cs with energies (eigenvalues) Es. The lowest energy eigenvalue represents an upper bound to the ground-state energy of the molecule; it would be the exact ground-state energy in a full CI calculation with an infinite number of spinorbitals. Provided the excited-state wavefunction is orthogonal to the ground-state wavefunction, the next lowest energy eigenvalue represents an upper bound for the first excited-state energy of the molecule, and so on. The matrix elements HIJ, which must be evaluated in CI calculations, can ultimately be expressed in terms of the basis functions y, because the Slater determinants are composed of spinorbitals expressed in terms of the basis functions (Fig. 9.7). When the number of determinants is large, there may be too many one- and two-electron integrals to store in the memory of the computer simultaneously, so their computation may have to be done in groups. In any event, conventional CI calculations are usually limited to a number of CSFs on the order of 104, and because full CI usually results in a list far exceeding this number, it is necessary to employ a truncation scheme so that the list of CSFs is kept at a manageable size. The use of a truncated CSF list is referred to as limited CI. A systematic approach to the selection of determinants for use in eqn 9.32 is to include all those determinants differing from the HF wavefunction F0 by no more than some predetermined number of spinorbitals. Because the spinorbitals cannot be improved once the HF-SCF calculation has been completed, the best we can do is systematically include more and more excited determinants in the expansion in eqn 9.31. Hamiltonian matrix elements H0J between the HF wavefunction F0 and determinants FJ that are more than doubly excited are zero. In addition, through the use of Brillouin’s theorem,16 hamiltonian matrix elements between F0 and all singly excited determinants also vanish. Thus, a first approach (which can be expected to be reasonably accurate when F0 is an approximation to the exact wavefunction) is to limit the list of excited determinants to those that are singly and doubly excited.17 Limitation of the list of determinants to F0 and determinants that are singly and doubly excited with respect to F0 is denoted SDCI. If only F0 and doubly excited determinants are used, then the technique is denoted DCI. Table 9.3 presents results for CI calculations on H2 using some of the basis sets of Table 9.1. In particular, we compare results from DCI and SDCI. As dihydrogen is a two-electron species, SDCI in this case is the same as full CI. The entries for the energies in Table 9.3 represent the differences between the SCF and CI ground-state energies, both computed using the same basis set. Several things are apparent from the table. First, the single excitations make a very small contribution to the energy. Second, the contribution to the .......................................................................................................
16. See Problem 9.17. 17. Singly excited determinants will have a small but non-zero effect on the calculation of the ground-state energy because they have non-zero matrix elements with doubly excited determinants, which themselves mix with F0. Moreover, single excitations do affect the electronic charge distribution and therefore properties such as the dipole moment. Thus, they are often included in CI calculations.
9.9 CI CALCULATIONS
j
307
Table 9.3 Calculated properties of dihydrogen
Basis set
{E(DCI) E(SCF)}/ (103 Eh)
{E(SDCI) E(SCF)}/ (103 Eh)
Re(SDCI)/a0
STO-3G
20.56
20.56
1.389
4-31G
24.87
24.94
1.410
6-31G
33.73
33.87
1.396
The energy differences (in hartrees, Table 9.1) are calculated at 1.4a0.
energy from the double excitations is very sensitive to the basis set. As the basis set gets larger, we recover a larger fraction of the exact correlation energy of 0.0409Eh (1.11 eV) from the doubly excited configurations. However, even the largest basis set of Table 9.3, 6-31G
, gives an energy that is significantly different from the exact correlation energy, primarily because l 2 functions have not been included in the basis. Table 9.3 also presents the equilibrium bond length of H2 obtained by finding the minimum in the calculated potential energy curve as a function of the bond length. We see that as the basis set is improved, the computed bond length is closer to the experimental result of 1.401a0. By comparing Tables 9.2 and 9.3, we can see that for a given basis set, the full CI calculation is superior to the HF-SCF result. Configuration interaction calculations that include single, double, triple, and quadruple excitations (in addition to F0) are designated SDTQCI. However, for basis sets large enough to recover most of the correlation energy, SDTQCI often involves too many determinants to be computationally practicable. As the quadruply excited determinants can be important in computing the correlation energy, a simple formula known as the Davidson correction has been proposed for estimating the contribution DEQ of quadruply excited determinants to the correlation energy:18
DEQ ¼ 1 C20 ðEDCI ESCF Þ ð9:38Þ where EDCI is the ground-state energy and C0 is the coefficient of F0 (for the normalized wavefunction of eqn 9.32), both obtained in a DCI calculation. ESCF is the ground-state energy associated with F0 obtained in an HF-SCF calculation. Using eqn 9.38 in a CI study of the ground state of N2, Langhoff and Davidson18 estimated that the contribution of quadruple excitations to the correlation energy of N2 was 7.6 per cent or 0.048hcR1 (corresponding to 24 millihartree, 0.65 eV). One serious deficiency that plagues limited CI calculations is the lack of size-consistency. A method is deemed size-consistent if the energy of a many-electron system is proportional to the number of electrons n in the limit of n ! 1; in particular, the energy of AB computed when subsystems A and B are infinitely far apart should be equal to the sum of the energies of .......................................................................................................
18. S.R. Langhoff and E.R. Davidson, Int. J. Quantum Chem., 61, 8 (1974).
308
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
A and B separately computed using the same method. That the physical requirement of size-consistency is not satisfied, for example, by a SDCI wavefunction is demonstrated in the following example. Example 9.3 Demonstrating the lack of size-consistency
Demonstrate that the SDCI calculation on the dimer He2 is not size-consistent. Method. To show that the calculation is not size-consistent, we need to show
that the energy of two infinitely separated He atoms is not equal to the energy of the dimer He2 when the internuclear distance in the dimer is infinite. Recalling that full CI is size-consistent but limited CI is not, we should compare SDCI calculations to full CI calculations on both He þ He and He2. Answer. First, consider the energies of the two He atoms separately computed using SDCI. The infinitely separated atoms each have a (restricted) HF wavefunction given by
F0 ¼ jjca1s ð1Þcb1s ð2Þjj Because He has only two electrons, a calculation involving all single and double excitations would involve all possible determinants: it would be a full CI calculation for each atom. Therefore, the SDCI calculation on the two independent two-electron He systems (that is, the four-electron He þ He SDCI calculation) includes contributions from quadruply excited determinants in which both electrons on each independent He atom are excited. Now consider the SDCI calculation on the composite four-electron He–He dimer with an infinite internuclear distance. An SDCI calculation involving only singly and doubly excited determinants will not be the same as full CI; in particular, it will not include contributions from quadruply excited determinants. Therefore, the SDCI calculation on the composite four-electron system He2 (at infinite internuclear separation) will result in both a different wavefunction and a different energy than the SDCI treatment of the two independent two-electron He systems. Thus, the limited CI calculation is not size-consistent. Self-test 9.3. What level of CI calculation is necessary to ensure that the He2
dimer CI calculation is size-consistent? [SDTQCI]
The magnitude of the size-consistency error increases as the size of the molecule increases. However, using the Davidson correction can reduce the error significantly.
9.10 Multiconfiguration and multireference methods In the CI methods described in the previous section, the expansion coefficients cji of eqn 9.14 are determined in an initial HF-SCF calculation and held fixed in the subsequent CI calculation. In the multiconfiguration self-consistent field method (MCSCF), the coefficients cji, as well as the coefficients CJs of eqn 9.32, are optimized. This simultaneous optimization of both sets of expansion coefficients makes MCSCF computationally demanding, but by optimizing cji,
9.10 MULTICONFIGURATION AND MULTIREFERENCE METHODS
j
309
accurate results can be obtained with the inclusion of a smaller number of CSFs. Development of efficient MCSCF methods is particularly important for excited states. One such scheme is the complete active-space self-consistent field method (CASSCF)19 in which the spinorbitals (which are themselves optimized during the calculation by determining the optimal values of cji) are divided into three classes: 1. A set of inactive orbitals composed of the lowest energy spinorbitals that are doubly occupied in all determinants included in eqn 9.32. 2. A set of virtual orbitals of very high energy spinorbitals that are unoccupied in all determinants. 3. A set of active orbitals that are energetically between the inactive and virtual orbitals. The active electrons are those electrons not in the doubly occupied inactive orbital set. The CSFs included in the CASSCF calculation are configurations (of the appropriate symmetry and spin) that arise from all possible ways of distributing the active electrons over the active orbitals. The choice of which orbitals to include as active orbitals is critical in CASSCF. One approach is to select the bonding, non-bonding, and antibonding orbitals that arise in qualitative MO theory from the valence atomic orbitals of the atoms in the molecule. For example, in a CASSCF calculation of the ground-state wavefunction and energy of the homonuclear diatomic B2, one choice could be to take the inactive orbitals to be the s-orbitals formed from the B1s atomic orbitals and to take the active orbitals to be the s- and p-orbitals that are formed from the 2s and 2p atomic orbitals. The choice of active orbitals is important because the number of CSFs rises very quickly as the number of active orbitals increases. In the restricted active-space (RAS) SCF method,20 the set of active orbitals is further divided into three subsets of orbitals (denoted I, II, and III) with the requirements that subset I contains a (specified) minimum number of electrons and subset III contains a (specified) maximum number of electrons. The number of electrons in subset II is unrestricted but the total number of electrons in the three subsets must be specified in the RASSCF calculation. In practice, subset II contains the most important orbitals for the problem at hand and this set of orbitals is reminiscent of the ‘normal’ active set of CASSCF. In the CI methods described so far, the HF-SCF wavefunction F0 is used as a reference configuration and configuration state functions are formed by moving electrons out of the occupied spinorbitals of F0 into unoccupied spinorbitals. In multireference configuration interaction (MRCI), a set of reference configurations is created, from which excited determinants are formed for use in a CI calculation. For example, one procedure would be to
.......................................................................................................
19. B.O. Roos, Int. J. Quantum Chem. Symp., 175, 14 (1980); L.M. Cheung, K.R. Sundberg, and K. Ruedenberg, Int. J. Quantum Chem., 1103, 16 (1979). 20. J. Olsen, B.O. Roos, P. Jørgensen, and H.J.A. Jensen, J. Chem. Phys., 2185, 89 (1988).
310
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
perform an MCSCF calculation and select a set of reference configurations from the determinants that have a coefficient CJs larger than some threshold value (such as 0.05 or 0.1) in the final normalized MCSCF wavefunction. For each reference determinant, electrons are moved from occupied spinorbitals to unoccupied spinorbitals to create more determinants for inclusion in the CI expansion in eqn 9.32. Then the configuration interaction calculation is performed, optimizing all the coefficients CJs of the determinants that have been included. The reference determinants will often be singly and doubly excited determinants with respect to F0 and single and double excitations from the reference determinants are often included. As a result, the final MRCI wavefunction will include determinants that are triply and quadruply excited from F0. Because MRCI calculations often include the most important quadruply excited determinants, they usually reduce the size-consistency error encountered in SDCI calculations significantly. In addition, it is often the case that a large fraction of the exact correlation energy can be recovered from MRCI calculations with a much smaller number of determinants.
9.11 Møller–Plesset many-body perturbation theory Configuration interaction calculations provide a systematic approach for going beyond the Hartree–Fock level, by including determinants that are successively singly excited, doubly excited, triply excited, and so on, from a reference configuration. One important feature of CI is that it is variational, but one disadvantage is its lack of size-consistency (with the exception of full CI). Perturbation theory (PT, Section 6.2) provides an alternative systematic approach to finding the correlation energy: whereas its calculations are sizeconsistent, they are not variational in that it does not in general give energies that are upper bounds to the exact energy. The application of PT to a system composed of many interacting particles is called many-body perturbation theory (MBPT). Because we want to find the correlation energy for the ground state, we take the zero-order hamiltonian from the Fock operators of the HF-SCF method. This choice of H(0) was made in the early days of quantum mechanics (in 1934) by C. Møller and M.S. Plesset, and the procedure is called Møller–Plesset perturbation theory (MPPT). Applications of MPPT to molecular systems did not actually begin until some 40 years later.21 In MPPT, the zero-order hamiltonian H(0) (in this context denoted HHF) is given by the sum of the one-electron Fock operators defined in eqn 9.9: HHF ¼
n X
fi
ð9:39Þ
i¼1
As we show in the following example, the HF ground-state wavefunction F0 ð0Þ is an eigenfunction of HHF with an eigenvalue E0 given by the sum of the orbital energies of all the occupied spinorbitals. .......................................................................................................
21. See J.A. Pople, J.S. Binkley, and R. Seeger, Int. J. Quantum Chem. Symp., 1, 10 (1976) for an early reference.
9.11 MØLLER–PLESSET MANY-BODY PERTURBATION THEORY
j
311
Example 9.4 Showing that F0 is an eigenfunction of HHF and determining its
eigenvalue Show that the HF ground-state wavefunction is an eigenfunction of the zeroorder MPPT hamiltonian and that its eigenvalue equals the sum of the occupied spinorbital energies. Method. Consider the hamiltonian HHF of eqn 9.39 and the effect of the one-
electron Fock operator fi on the spinorbital fa (eqn 9.8). Analyse the effect of HHF on each term in the expansion of F0. Use the fact that a linear combination of eigenfunctions of an hermitian operator all having an identical eigenvalue is itself an eigenfunction with the same eigenvalue. Answer. The HF ground-state wavefunction is
F0 ¼ fa ð1Þfb ð2Þ . . . fz ðnÞ
An expansion of the Slater determinant yields a sum of terms, each of which involves a product of the n spinorbitals fafb . . . fz with the electrons 1, 2, . . . , n distributed differently in each term in the summation. Consider the effect of HHF on one of these terms, the principal diagonal of the determinant, using eqns 9.39 and 9.8: n X HHF fa ð1Þfb ð2Þ . . . fz ðnÞ ¼ fi fa ð1Þfb ð2Þ . . . fz ðnÞ i¼1
¼ ff1 fa ð1Þgfb ð2Þ . . . fz ðnÞ þ fa ð1Þff2 fb ð2Þg . . . fz ðnÞ þ þ fa ð1Þfb ð2Þ . . . fn fz ðnÞ ¼ ðea þ eb þ þ ez Þfa ð1Þfb ð2Þ . . . fz ðnÞ Each term in the expansion of F0 has each occupied spinorbital appearing once; therefore each term in the expansion is an eigenfunction of HHF with the same eigenvalue ea þ eb þ þ ez. We can immediately conclude that F0 is an eigenfunction of the MPPT zero-order hamiltonian with an eigenvalue given by the sum of the orbital energies of all occupied spinorbitals. Self-test 9.4. Show that all singly and multiply excited determinants are also
eigenfunctions of HHF with eigenvalues equal to the sums of the orbital energies of the spinorbitals occupied in that particular determinant.
The perturbation H(1) is given by n X fi Hð1Þ ¼ H
ð9:40Þ
i¼1
where, as before, H is the electronic hamiltonian. The HF energy EHF associated with the (normalized) ground-state HF wavefunction F0 is the expectation value ð9:41Þ EHF ¼ hF0 jHjF0 i or, equivalently, EHF ¼ hF0 jHHF þ Hð1Þ jF0 i
ð9:42Þ
312
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE ð0Þ
It is easy to show that EHF is equal to the sum of the zero-order energy E0 ð1Þ and the first-order energy correction E0 . From eqns 6.12 and 6.20 and the fact that F0 is an eigenfunction of HHF, we know ð0Þ
ð9:43Þ
ð1Þ
ð9:44Þ
E0 ¼ hF0 jHHF jF0 i E0 ¼ hF0 jHð1Þ jF0 i From eqns 9.42–9.44, we conclude that ð0Þ
ð1Þ
EHF ¼ E0 þ E0
Therefore, the first correction to the ground-state energy is given by secondorder perturbation theory as (see eqn 6.24) X hFJ jHð1Þ jF0 ihF0 jHð1Þ jFJ i ð9:45Þ Eð2Þ ¼ ð0Þ ð0Þ E0 EJ J6¼0 where FJ is a multiply excited determinant and an eigenfunction of HHF with ð0Þ eigenvalue EJ . To evaluate eqn 9.45, we need to be able to evaluate the offdiagonal matrix elements hFJjH(1)jF0i. First, we note that the matrix element hFJ jHHF jF0 i ¼ 0 because F0 is an eigenfunction of HHF and the spinorbitals, and hence the determinants, are orthogonal. Therefore, if hFJ jHjF0 i ¼ 0,
then hFJ jHð1Þ jF0 i ¼ 0
From Brillouin’s theorem and the discussion in Section 9.9, we conclude that only the doubly excited determinants have non-zero H(1) matrix elements with F0 and therefore only double excitations contribute to E(2). An analysis of these non-vanishing matrix elements22 yields the following expression: Eð2Þ ¼ 14
occ X vir X ðabjjpqÞðpqjjabÞ a;b p;q
where
ea þ eb ep eq
Z
1 f a ð1Þf b ð2Þ f ð1Þfq ð2Þ dx1 dx2 r12 p Z 1 j0 f a ð1Þf b ð2Þ f ð1Þfp ð2Þ dx1 dx2 r12 q
ðabjjpqÞ ¼ j0
ð9:46Þ
ð9:47Þ
with fa and fb being occupied spinorbitals and fp and fq being virtual spinorbitals. The inclusion of the second-order energy correction in MPPT is designated MP2. In general, bond lengths computed using MP2 are in excellent agreement with experiment for bonds involving hydrogen. However, the same is not generally true for multiple bonds. For example, the bond lengths for N2 from MP2 are 2.322a0 (STO-3G basis), 2.171a0 (4-31G basis), and 2.133a0 (6-31G
basis) compared to the experimental value of 2.074a0. .......................................................................................................
22. See, for example, Section 2.3.4.3 of D.M. Hirst, A computational approach to chemistry, Blackwell Scientific Publications, Oxford (1990).
9.12 THE COUPLED-CLUSTER METHOD
j
313
It is possible to extend MPPT to include third- and fourth-order energy corrections, and the procedures are then denoted MP3 and MP4.23 The algebra involved becomes more complicated at higher orders of perturbation theory, and it is common to use diagrammatic techniques to classify and represent the various terms that appear in the perturbation series expressions. These diagrammatic representations can be used to prove that MPPT is sizeconsistent in all orders.
9.12 The coupled-cluster method Another popular ab initio method that, like MPPT, is size-consistent but not variational, is called the coupled-cluster method (CC method). The CC method introduces the cluster operator C, which relates the exact electronic wavefunction C to the HF wavefunction F0 through C ¼ e C F0 where the exponential operator eC is defined by the series expansion eC ¼ 1 þ C þ 2!1 C2 þ 3!1 C3 þ We show below that the effect of eC on F0 is to create a linear combination of Slater determinants that includes F0 and all its singly, doubly, . . . , N-tuply excited determinants (reminiscent of eqn 9.31). Therefore, provided the spinorbitals comprising F0 form a complete basis set (which, of course, in practice will never be the case), C is the exact wavefunction for any state of the system. The effect of the cluster operator C on F0 is to give a linear combination of Slater determinants in which electrons from occupied spinorbitals have been excited to virtual spinorbitals. In particular, C is the sum of the oneelectron excitation operator C1, two-electron excitation operator C2, . . . , N-electron excitation operator CN: C ¼ C1 þ C2 þ þ CN The effects of the excitation operators are X C1 F0 ¼ tap Fap a;p
C2 F0 ¼
X
pq pq tab Fab
a;b;p;q
and likewise for C3 to CN; the tap are called single-excitation amplitudes, pq double-excitation amplitudes, and so on. No operators beyond CN appear tab because F0 has all electrons in N occupied spinorbitals. The excitation amplitudes are determined by solving the coupled cluster equations; the latter set of equations is derived by substituting eCF0 into the electronic Schro¨dinger equation.24 .......................................................................................................
23. For a detailed discussion, see S. Wilson, Electron correlation in molecules, Clarendon Press, Oxford (1984). 24. For details, see F. Jensen, An introduction to computational chemistry, Wiley, Chichester (1999). A useful review article on CC is R.J. Bartlett, J. Phys. Chem., 1697, 93 (1989).
314
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
The effect of eC on F0 yields terms of the form C1F0, C2F0, C3F0, . . . , but it also results in products of excitation operators such as C1C1F0, C1C2F0, and C1C2C3F0. Because C1F0 results in singly excited determinants, another application of C1 (as in C1C1F0) results in doubly excited determinants; the latter also result from C2F0. However, there is an important difference. pq appear whereas for C1C1F0, For C2F0, the double-excitation amplitudes tab p q products ta tb of single-excitation amplitudes result. We say that C2F0 represents a ‘connected’ double-excitation contribution whereas C1C1F0 represents a ‘disconnected’ double-excitation contribution.25 Similarly, C3F0 is a connected triple-excitation contribution while C1C1C1F0 (that is, C13F0) and C1C2F0 are disconnected, involving, respectively, products of three single-excitation amplitudes and products of single- and double-excitation amplitudes. Only a finite number of terms will appear in the Taylor-series expansion of eCF0 as many terms, such as C1N þ 1F0, vanish. Two approximations are widely made in CC applications. First, as discussed previously, a finite (and hence incomplete) basis set is used in the determination of F0. Second, the expression for the cluster operator C is truncated to include only specified electron excitation operators. In the approach referred to as ‘coupled cluster singles and doubles’ (CCSD), C is approximated by C1 þ C2. In CCD, only C2 is employed whereas in CCSDT, C is given by C1 þ C2 þ C3. Example 9.5 Deriving the set of coupled-cluster equations
Derive the set of coupled-cluster equations for the double-excitation amplitudes in the CCD method. Method. Substitute the expression eCF0 into the electronic Schro¨dinger
equation (taking the cluster operator C in the CCD approach to be C2); follow that by multiplication of both sides of the equation by F j (where FJ is the HF wavefunction F0 or any excited Slater determinant) and integration over all spin–space coordinates. Use the fact that hamiltonian matrix elements between Slater determinants differing by four (or more) spin orbitals are zero. (The latter result is an example of a Slater–Condon rule, Problem 9.18.) Answer. Substitution of eCF0 into the electronic Schro¨dinger equation yields
HeC F0 ¼ EeC F0 and using the series expansion of eC with C ¼ C2 gives Hð1 þ C2 þ 12C22 þ ÞF0 ¼ Eð1 þ C2 þ 12C22 þ ÞF0 Inserting the effects of the excitation operator C2 gives X pq pq X pq pqrs HF0 þ H tab Fab þ 12 H trs cd t ab Fabcd þ a;b;p;q
¼ EF0 þ E
X a;b;p;q
a;b;c;d; p;q;r;s
tpq ab
Fpq ab
þ 12 E
X
pq pqrs trs cd t ab Fabcd þ
a;b;c;d; p;q;r;s
.......................................................................................................
25. The words connected and disconnected are intimately related to linked and unlinked features in the diagrammatic representation of the coupled-cluster method.
9.12 THE COUPLED-CLUSTER METHOD
j
315
We multiply both sides of the above equation by F 0 and integrate over spin– space coordinates. Because F0 is normalized and orthogonal to all excited Slater determinants, and because hamiltonian matrix elements between Slater determinants differing by four (or more) spinorbitals are zero, we obtain X pq EHF þ tab hF0 jHjFpq ab i ¼ E a;b;p;q
where the HF-SCF energy EHF ¼ hF0jHjF0i. Proceeding in the same way we did above for F 0 , we now multiply both sides of the Schro¨dinger equation by the doubly excited Fkl
ij and integrate over spin– space coordinates. Using the orthonormality of the Slater determinants and the Slater–Condon rules, we find X X pq pq pqrs rs pq 1 hFkl tab hFkl tcd tab hFkl ij jHjF0 i þ ij jHjFab i þ 2 ij jHjFabcd i a;b; p;q
¼E
X
a;b;c;d; p;q;r;s pq pq tab hFkl ij jFab i
a;b; p;q
¼ Etijkl Finally, inserting the expression for E given above (which was obtained from multiplication of the Schro¨dinger equation by F 0 ), we obtain X X pq pq pqrs rs pq 1 hFkl tab hFkl tcd tab hFkl ij jHjF0 i þ ij jHjFab i þ 2 ij jHjFabcd i a;b; p;q
( ¼
EHF þ
a;b;c;d; p;q;r;s
X
)
pq tab hF0 jHjFpq ab i
tijkl
a;b; p;q
If there is a total of m doubly excited determinants Fkl ij , there are m equations of the form given above and there are m unknown double-excitation amplitudes tijkl . Once the hamiltonian matrix elements are computed (as well as the HF energy), the set of non-linear equations for the double-excitation amplitudes is solved (usually in an iterative fashion beginning with estimates of the amplitudes). From the amplitudes the CCD energy E and wavefunction eCF0 are determined. Self-test 9.5. Derive the set of coupled-cluster equations for the excitation
amplitudes in the CCSD method.
We conclude this section by making some comparisons between the CI and CC methods, both of which are configuration-state-function approaches to electronic structure. For a specified basis set for determination of the spinorbitals, a full CI calculation and a CC calculation including all excitation operators (C1, C2, . . . , CN) would yield identical electronic energies. In the CCD approach, the wavefunction eCF0 includes the HF wavefunction and all its doubly excited, quadruply excited, hextuply excited, . . . , determinants. Only the double-excitations are connected; for example, the contributions from the quadruply excited determinants are given by (disconnected)
316
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
products of two double-excitation amplitudes. Therefore, quadruply excited determinants are not treated exactly as they are in SDTQCI; in the latter, the pqrs doubly and quadruply excited determinants have coefficients Cpq ab and Cabcd that are determined independently. However, it is often the case that the CCD term C22 F0 is able to account for most of the effects of quadruply excited pq rs determinants; in other words, approximating Cpqrs abcd as products tab tcd is often quite reasonable. CCD is generally more accurate than DCI but requires more computer time; however, this CC method requires significantly less computer time than (the slightly more accurate but not size-consistent) SDTQCI. The CCSD approach is generally more popular currently than CCD because the inclusion of the one-electron excitation operator has little effect on the overall computation time.
Density functional theory
Our use of the word functional refers to a mathematical prescription for assigning a number to a given function. As an example of a functional in classical mechanics, suppose a particle moves in the xy plane from point C to point D with velocity v(x,y). Then the time the particle takes to move from C to D is a functional of the velocity. A good reference on the theory of functionals is I.M. Gelfand and S.V. Fomin (revised English edition translated and edited by R.A. Silverman), Calculus of variations, Prentice-Hall, Inc. Englewood Cliffs, New Jersey (1963).
The ab initio methods described above all start with the Hartree–Fock approximation in that the HF equations are first solved to find spinorbitals that can then be used to construct configuration state functions. These methods are widely used by quantum chemists today. However, they do have limitations, in particular the computational difficulty of performing accurate calculations with large basis sets on molecules containing many atoms and many electrons. An alternative to the HF methods that is also popular among quantum chemists is density functional theory (DFT). In contrast to the methods described above, which use CSFs, DFT begins with the concept of the electron probability density. One reason for the popularity of DFT is that it takes into account electron correlation while being less demanding computationally than, for example, CI and MP2. It can be used to do calculations on molecules of 100 or more atoms in significantly less time than these HF methods. Furthermore, for systems involving d-block metals, DFT yields results that very frequently agree more closely with experiment than HF calculations do. The basic idea behind DFT is that the energy of an electronic system can be written in terms of the electron probability density, r.26 For a system of n electrons, r(r) denotes the total electron density at a particular point r in space. The electronic energy E is said to be a functional of the electron density and is denoted E[r], in the sense that for a given function r(r), there is a single corresponding energy.27 .......................................................................................................
26. For more extensive treatments, the reader is encouraged to see the qualitative discussion in S. Borman, Chem. Eng. News, 22, 68 (1990) and a more detailed quantitative discussion in the review by T. Ziegler, Chem. Rev., 651, 91 (1991). A review by W. Kohn can be found in Rev. Mod. Phys., 1253, 71 (1999). More recent reviews include M.K. Harbola and A. Banerjee, J. Theor. Comput. Chem., 301, 2 (2003) and T. Nakajima, T. Tsuneda, H. Nakano, and K. Hirao, J. Theor. Comput. Chem., 109, 1 (2002). 27. We have encountered the functional before, but we did not use this name. An expectation value of a hamiltonian is the energy as a functional of the wavefunction, c; each well-behaved function c is associated with a single expectation value of the energy.
9.13 KOHN–SHAM ORBITALS AND EQUATIONS
j
317
9.13 Kohn–Sham orbitals and equations The concept of a density functional for the energy was the basis of some early but useful approximate models such as the Thomas–Fermi method (which emerged from work in the late 1920s by E. Fermi and L.H. Thomas) and the Hartree–Fock–Slater or X method (which emerged from the work of J.C. Slater in the 1950s). However, it was not until 1964 that a formal proof was given28 by P. Hohenberg and W. Kohn that the ground-state energy and all other ground-state electronic properties are uniquely determined by the electron density. Unfortunately, the Hohenberg–Kohn theorem does not tell us the form of the functional dependence of energy on the density: it proves only that such a functional exists. The next major step in the development of DFT came with the derivation of a set of one-electron equations from which the electron density r could be obtained.29 We consider systems in which paired electrons are described by the same spatial one-electron orbitals (as in restricted Hartree–Fock theory). W. Kohn and L.J. Sham showed that the exact ground-state electronic energy E of an nelectron system can be written as E½r ¼
n 2 X h 2me i¼1
þ 12j0
Z
Z
c i ðr 1 Þr21 ci ðr 1 Þ dr 1 j0
N X ZI
r I¼1 I1
rðr 1 Þ dr 1 ð9:48Þ
rðr 1 Þrðr 2 Þ dr 1 dr 2 þ EXC ½r r12
where the one-electron spatial orbitals ci (i ¼ 1, 2, . . . , n) are the Kohn–Sham orbitals, the solutions of the equations given below. The exact ground-state electron density is given by rðrÞ ¼
n X
jci ðrÞj2
ð9:49Þ
i¼1
where the sum is over all the occupied Kohn–Sham (KS) orbitals; r is known once these orbitals have been computed. The first term on the right in eqn 9.48 represents the kinetic energy of the electrons; the second term represents the electron–nucleus attraction where the sum is over all N nuclei with index I and atomic number ZI; the third term represents the Coulomb interaction between the total charge distribution (summed over all KS orbitals) at r1 and r2; the last term is the exchange–correlation energy of the system, which is also a functional of the density and takes into account all non-classical electron– electron interactions. Of the four terms, EXC is the one we do not know how to obtain exactly. Although the Hohenberg–Kohn theorem tells us that E and therefore EXC must be functionals of the electron density, we do not know the latter’s exact analytical form and so are forced to use approximate expressions for it. .......................................................................................................
28. P. Hohenberg and W. Kohn, Phys. Rev., 864, B136 (1964). 29. W. Kohn and L.J. Sham, Phys. Rev., 1133, A140 (1965).
318
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
The KS orbitals are found by solving the Kohn–Sham equations, which are derived by applying the variational principle to the electronic energy E[r] with the charge density given by eqn 9.49. The KS equations for the oneelectron orbitals ci(r1) have the form (
N X 2 2 h ZI r1 j0 þ j0 2me r I¼1 I1
Z
) rðr 2 Þ dr 2 þ VXC ðr 1 Þ ci ðr 1 Þ ¼ ei ci ðr 1 Þ r12 ð9:50Þ
Consider a functional G[f] that depends on the function f(r). When r undergoes an arbitrarily small change dr and the function changes to f þ df, the functional undergoes a corresponding change to G[f þ df]. We then define the functional derivative dG/df as dG G½f þ df G½f ¼ lim df jdf j jdf j!0 where the manner in which jdfj ! 0 must be specified explicitly.
where ei are the KS orbital energies and the exchange–correlation potential, VXC, is the functional derivative of the exchange–correlation energy: VXC ½r ¼
dEXC ½r dr
ð9:51Þ
If EXC is known, then VXC can be obtained. The significance of the KS orbitals is that they allow the density r to be computed from eqn 9.49. The KS equations are solved in a self-consistent fashion. Initially, we guess the electron density r, typically by using a superposition of atomic densities. By using some approximate form (which remains fixed during all iterations) for the functional EXC[r], we next compute VXC as a function of r. The set of KS equations is then solved to obtain an initial set of KS orbitals. This set of orbitals is then used to compute an improved density from eqn 9.49, and the process is repeated until the density and exchange–correlation energy have converged to within some tolerance.30 The electronic energy is then computed from eqn 9.48. The KS orbitals can be computed numerically or they can be expressed in terms of a set of basis functions; in the case of the latter, solving the KS equations amounts to finding the coefficients in the basis set expansion. As in the HF methods, a variety of basis set functions can be used (including STOs and GTOs) and the wealth of experience gained in HF calculations can prove to be useful in the choice of DFT basis sets. The computation time required for a DFT calculation formally scales as the third power of the number of basis functions; as a result, DFT methods are computationally more efficient (though not necessarily more accurate) than HF-based formalisms, which scale as the fourth power of the number of basis functions. However, for large systems such as proteins, even this thirdpower scaling makes computational investigations impractical and much effort these days is devoted to the development of DFT algorithms with lower-power scaling.31
.......................................................................................................
30. Often, using the newly found density to start the next iteration will not lead to convergence but rather to oscillatory behaviour. One simple procedure to enhance convergence is to use, for iteration i þ 2, the average of the densities from iterations i and i þ 1. 31. See G. te Velde, F.M. Bickelhaupt, E.J. Baerends, C.F. Guerra, S.J.A. van Gisbergen, J.G. Snijders, and T. Ziegler, J. Comput. Chem., 931, 22 (2001) and references therein.
9.14 EXCHANGE–CORRELATION FUNCTIONALS
j
319
9.14 Exchange–correlation functionals Numerous schemes have been developed for obtaining approximate forms for the functional for the exchange–correlation energy; the search for more accurate functionals is an active area of current research efforts. The main source of error in DFT usually stems from the approximate nature of EXC. This functional is often separated into an exchange functional (representing exchange energy) and a correlation functional (representing dynamic correlation energy). In the local density approximation (LDA), the exchange– correlation functional is Z ð9:52Þ EXC ¼ rðrÞeXC ½rðrÞ dr
Table 9.4 Calculated and experimental metal–ligand mean bond energies
Calculated
Observed
Cr(CO)6
107
110
Mo(CO)6
126
151
W(CO)6
156
179
Energies are in kilojoules per mole of M–L bonds (kJ mol1).
where eXC[r(r)] is the exchange–correlation energy per electron in a homogeneous electron gas of constant density. In a hypothetical homogeneous electron gas, an infinite number of electrons travel throughout a space of infinite volume in which there is a uniform and continuous distribution of positive charge to retain electroneutrality.32 Although eqn 9.52 is clearly an approximation (neither positive charge nor electronic charge is uniformly distributed in actual molecules), the computationally convenient LDA is surprisingly accurate, especially for predicting structural properties. The accuracy decreases with varying electron density in the system and for many molecules, the LDA (for which the exchange functional and correlation functional depend only on r but not on any derivatives of r) was found to yield values for binding energies that were significantly larger than experiment. To account for the inhomogeneity of the electron density, a non-local correction involving the gradient of r is often added to the exchange– correlation energy given in eqn 9.52. A number of different gradient-correct functionals have been proposed; in general, the LDA with gradient corrections, which is called the generalized gradient approximation (GGA), yields ground-state bond distances accurate to within 0.03a0 (1.5 pm) and binding energies accurate to within about 20 kJ mol1. The GGA-DFT procedure is an accurate and efficient method for calculations involving d-metal complexes. Table 9.4 compares calculated and experimental values of M–CO bond strengths for several d-block metals. A variety of exchange–correlation functionals, having names such as mPWPW91, B3LYP, MPW1K, PBE1PBE, BLYP, BP91, and PBE, have been developed for use in DFT calculations; the names designate a particular pairing of an exchange functional and a correlation functional. For example, the popular BLYP functional is a combination of the gradient-corrected exchange functional developed by A.D. Becke33 and the gradient-corrected
.......................................................................................................
32. Several forms for eXC[r] have been proposed in the literature; an accurate expression can be found in R.G. Parr and W. Yang, Density-functional theory of atoms and molecules, Oxford University Press, Oxford (1989). 33. A.D. Becke, J. Chem. Phys., 4524, 84 (1986); Phys. Rev. A, 3098, 38 (1988).
320
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
correlation functional developed by C. Lee, W. Yang, and R.G. Parr.34 Some of the functionals, such as B3LYP, represent ‘hybrid’ DFT calculations that use Hartree–Fock corrections in conjunction with density functional correlation and exchange. In a recent publication,35 a variety of DFT-based calculations were performed to compute barrier heights for six smallmolecule reactions and atomization (complete dissociation) energies for six different molecules. The results for a variety of DFT functionals and basis sets were compared with each other and with results of Hartree–Fock-based techniques. Density functional theory has also been applied to the study of open-shell atoms and molecules. The extension of LDA to open-shell systems yields the local spin-density approximation (LSDA). The spin density refers to the difference between the spin-up electron density and the spin-down electron density, and in the LSDA the exchange–correlation energy depends on the spin density as well as the total electron density. The LSDA has been used in DFT investigations of the magnetic structures of metals and alloys. Time-dependent density functional theory (TD-DFT) is useful for the investigation of the response of molecular systems to electric and magnetic fields (recall the discussion of time-dependent hamiltonians in Chapter 6). Therefore, TD-DFT can be used to determine polarizabilities and hyperpolarizabilities (Chapter 12) as well as excitation energies and electronic absorption spectra. The time-dependent Kohn–Sham equations in TD-DFT resemble those of eqn 9.50: (
) Z N X 2 2 h ZI rðr 2 , tÞ r j0 þ j0 dr 2 þ Vext ðtÞ þ VXC ðr 1 , tÞ ci ðr 1 , tÞ r12 2me 1 r I¼1 I1
q c ðr 1 , tÞ qt i n X jci ðr, tÞj2 rðr, tÞ ¼ ¼ ih
i¼1
where the external potential Vext (for example, the oscillating electromagnetic field), the exchange–correlation potential, the KS orbitals and the density are all time-dependent. Although the functional dependence of VXC(r, t) on r(r, t) need not match the functional dependence of the time-independent exchange– correlation potential on the time-independent density (of conventional DFT), the same functional dependence is often assumed. The latter assumption is called the adiabatic approximation. The goal of TD-DFT is to find how the density changes in response to the varying external potential. It can be shown, based on perturbation theory arguments, that excitation energies and polarizabilities depend on only the first-order change in density.36 .......................................................................................................
34. C. Lee, W. Yang, and R.G. Parr, Phys. Rev. B, 785, 37 (1988). 35. B.J. Lynch and D.G. Truhlar, J. Phys. Chem., 8996, 107 (2003). 36. Interested readers should see the reference cited in footnote 31 as well as the Further reading.
9.15 ENERGY DERIVATIVES AND THE HESSIAN MATRIX
j
321
Gradient methods and molecular properties
The gradient of a scalar function f(x,y,z) is defined as
qf qf qf iþ jþ k rf ¼ qx qy qz where i, j, and k are unit vectors in the x-, y-, and z-directions, respectively. Some properties of the gradient are given in Further information 22.
Once the electronic energy is obtained by solving the electronic Schro¨dinger equation, a number of molecular properties, perhaps the most important being the equilibrium molecular geometry, can be determined. The calculation of molecular structures is a valuable supplement to experimental data in areas of structural chemistry such as X-ray crystallography, electron diffraction, and microwave spectroscopy. Calculation of derivatives of the potential energy with respect to nuclear coordinates is crucial to the efficient determination of equilibrium structures. The derivatives can be computed numerically by calculating the potential energy at many geometries and determining the change in energy as each nuclear coordinate is varied. However, gradient methods, which determine energy derivatives analytically, are computationally faster and more accurate than numerical differentiation.
9.15 Energy derivatives and the Hessian matrix Since 1969, when P. Pulay wrote the first computer program for determining first derivatives of SCF energies analytically, gradient methods have developed into one of the most vigorously studied areas of modern quantum chemistry. First applied to closed-shell SCF calculations, gradient methods were later generalized to open-shell RHF and UHF calculations. In addition to the development of gradient methods for ab initio techniques based on Slater determinants, analytical expressions have also been derived for DFT. In general, analytical first and second energy derivatives are now available for a number of levels of ab initio calculations. For a diatomic molecule, the molecular potential energy, E, depends only on the internuclear distance, R; therefore, to find the potential minimum (more generally, any stationary point) we need to locate a zero in dE/dR.37 The search is more complicated for polyatomic molecules because the potential energy is a function of many nuclear coordinates, qi. At the equilibrium geometry, each of the forces fi exerted on a nucleus by electrons and other nuclei must vanish: fi ¼
qE ¼0 qqi
Therefore, in principle, the equilibrium geometry can be found by computing all the forces at a given molecular geometry and seeing if they vanish. If they do not, the geometry is varied until one is found that corresponds to zero forces, a gradient vector of zero length. Computationally, the forces will not vanish identically, but we can stop the iterative search for the equilibrium geometry when the magnitudes of the forces are sufficiently close to zero (that is, when the magnitudes are smaller than a predetermined tolerance level). Typically, .......................................................................................................
37. In this section, the nucleusnucleus repulsion term is included in the potential energy.
322
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
to obtain bond and torsional angles and bond distances within 1 and 0.002a0 (0.1 pm), respectively, of their optimal computed equilibrium values, all forces should be below about 10 pN. Such tolerances typically require between N and 2N cycles, where N is the number of atoms in the molecule. Before optimizing the equilibrium geometry, a coordinate system must be chosen to represent the potential surface and molecular structure. This choice is important because it affects the ease of optimization, and internal coordinates (bond lengths, bond angles, and torsional angles) are often used. Additionally, it is important to recognize that experimental data usually refer to averaged molecular quantities. Therefore, when comparing experimental and theoretical molecular geometries, vibrationally averaged bond lengths and bond angles should be computed. The differences between experimental and computed equilibrium geometries for a series of related compounds are often found to be very consistent. For instance, in an HF-SCF study of 30 organic compounds using a 4-21G basis set, the optimized C–H bond distance was consistently about 0.07a0 smaller than the experimental value.38 This consistency makes it possible in many cases to construct a set of empirical corrections for ab initio geometries yielding absolute accuracies of about 0.02a0. A zero gradient characterizes a stationary point on the surface but does not differentiate between minima, maxima, and saddle points.39 Therefore, the searching procedure allows us to locate not only an equilibrium geometry of a stable molecule but also the transition state of a chemical reaction, the latter corresponding to a saddle point on the potential energy surface. To distinguish the types of stationary points, it is necessary to consider the second derivatives of the energy with respect to the nuclear coordinates. The quantities q2E/qqi qqj comprise the Hessian matrix. Whereas a minimum (maximum) of a one-dimensional potential curve corresponds to a positive (negative) second derivative, a minimum (maximum) of a multidimensional potential energy surface is characterized by the eigenvalues of the Hessian matrix all being positive (negative). A transition state (a first-order saddle point) corresponds to one negative eigenvalue and all the rest positive.
9.16 Analytical derivatives and the coupled perturbed equations There are a number of algorithms available for finding stationary points on a potential surface. In general, the stability, reliability, and computational cost of the algorithm as well as its speed of convergence need to be considered. The algorithms can be broadly classified into three groups. Those using only the energy are the slowest to converge but are useful if analytical derivatives are unavailable. Those using both the energy and its analytical first derivatives
.......................................................................................................
38. An extensive comparison of theory and experiment is made in W.J. Hehre, L. Radom, P.V.R. Schleyer, and J.A. Pople, Ab initio molecular orbital theory, Wiley, New York (1986). 39. A (first-order) saddle point is a potential maximum along one nuclear coordinate and a potential minimum along all others.
9.16 ANALYTICAL DERIVATIVES AND THE COUPLED PERTURBED EQUATIONS
j
323
are significantly more efficient (by almost an order of magnitude). Furthermore, their rate of convergence can be improved if a good initial estimate of the Hessian matrix is available, perhaps obtained from lower level ab initio calculations (for example, ones that use smaller basis sets). Algorithms that use the energy together with its analytical first and second derivatives are the most accurate and efficient methods. Whichever algorithm is used, all nuclear coordinates should be optimized; this optimization is especially important for transition states where optimizing a subset of all the nuclear coordinates might locate a saddle point that changes significantly when all coordinates are optimized. Energy derivatives are also useful for determining other molecular properties. The second derivatives of the energy with respect to nuclear coordinates (the Hessian matrix elements) are the force constants for normal mode frequencies within the harmonic approximation (Section 10.13). The third, fourth, and higher derivatives give anharmonic corrections to vibrational frequencies (Section 10.16). Energy derivatives need not be limited to nuclear coordinates; for example, it is sometimes useful to consider derivatives with respect to electric field components. Mixed second derivatives with respect to one nuclear coordinate and one electric field component yield dipole moment derivatives that are used to determine infrared intensities within the harmonic approximation (see Section 10.10). To calculate analytical derivatives of the energy with respect to nuclear coordinates it is necessary to compute derivatives of one- and two-electron integrals over the basis functions. Because the basis functions are centred on atomic nuclei, when derivatives of the integrals are determined we need the derivatives of the basis set functions with respect to nuclear coordinates. Whether or not derivatives of various expansion coefficients are also required depends on whether they were determined variationally and on the order of the energy derivative under consideration. As an example, consider derivatives of a general expansion coefficient denoted cj. The first derivative of the energy with respect to nuclear coordinate qi is given by dE qE X qE qcj ¼ þ dqi qqi qcj qqi j However, for variationally determined expansion coefficients, the term (qE/qcj) vanishes; therefore, we have the important result that to evaluate the gradient (first derivative) of the energy, we do not need derivatives of variationally determined coefficients. As a result, in HF and MCSCF, the analytical determination of the energy gradient requires derivatives of only the one- and two-electron integrals. Similarly, the second derivative of the energy is given by ! d2 E q2 E X q2 E qcj qE q2 cj ¼ 2þ þ qqi qcj qqi qcj qq2i dq2i qqi j As the term (qE/qcj) vanishes, evaluation of the second derivative of the energy does not require the second derivative of the variationally determined
324
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
coefficients. However, it does require their first derivatives, (qcj/qqi), which are computed by solving the coupled perturbed Hartree–Fock equations (CPHF) for those ab initio methods that use a single reference CSF (for instance, SDCI and MP2) or by solving the coupled perturbed MCSCF equations (CPMCSCF) for those using multiple reference CSFs (for instance, MCSCF and MRCI).40 In general, to compute energy derivatives of order (2n þ 1) requires derivatives of variationally determined coefficients of order n.41 Therefore, the third derivative of the energy requires first derivatives of variational coefficients. An efficient method for solving the CPHF equations was first developed by J.A. Pople and co-workers in 1979 that made energy second-derivative calculations practicable for RHF and UHF. Analytical first derivatives of the energy do require, however, determination of the first derivatives of non-variationally determined coefficients. The latter derivatives are obtained by solving the appropriate coupled perturbed equations. However, a significant simplification of the solution of the coupled perturbed equations was discovered by N.C. Handy and H.F. Schaefer in 1984, who showed that instead of solving all the CPHF or CPMCSCF equations, it is in fact necessary to solve only a much smaller set of equations. This simplification is also applicable to higher energy derivatives; for example, when determining second derivatives of the energy, the full set of coupled perturbed equations for the second derivatives of the non-variational coefficients can also be reduced to a smaller set. To calculate analytical derivatives of the energy, it is necessary to evaluate the derivatives of the one- and two-electron integrals; this in turn requires evaluation of derivatives of the basis set functions with respect to the nuclear coordinates of their centres. For Gaussian-type orbitals (eqn 9.28), the derivatives of the basis functions may be computed analytically and result in other GTOs. For example, the first derivative of an s-type Gaussian with respect to xc yields a p-type GTO; the first derivative of the p-type GTO y100 yields an s- and a d-type GTO (see Problem 9.25).
Example 9.6 Analytically computing the derivatives of Gaussian-type orbitals
Show that the first derivative an of s-type Gaussian with respect to one of the Cartesian coordinates yields a p-type Gaussian. Method. The general form of a GTO is given in eqn 9.28. Differentiate y000
with respect to the Cartesian coordinate xc and compare the result to the p-type Gaussian y100.
.......................................................................................................
40. When the expansion coefficients are constrained, for example by orthonormality considerations, appropriate Lagrangian multipliers must be introduced into the coupled perturbed equations. 41. The analogy to perturbation theory (Section 6.6) should be noted. In fact, the connection between PT and gradient methods runs deeper than may be apparent; early investigations of SCF derivatives were often couched in the language of PT.
9.16 ANALYTICAL DERIVATIVES AND THE COUPLED PERTURBED EQUATIONS
j
325
Answer. The s-type GTO is given by eqn 9.28 with i ¼ j ¼ k ¼ 0. As jr1 rcj2 ¼
(x1 xc)2 þ (y1 yc)2 þ (z1 zc)2, 2
2
2
y000 ¼ eaðx1 xc Þ eaðy1 yc Þ eaðz1 zc Þ
and differentiation with respect to xc yields qy000 ¼ 2aðx1 xc Þy000 ¼ 2ay100 qxc where y100 is a p-type GTO. Self-test 9.6. What result is obtained if the first derivative of the s-type GTO is
taken with respect to yc or zc? [2ay010 or 2ay001]
Much work has gone into the efficient evaluation of integral derivatives (see Further reading). The efficiency can be increased by using translational and rotational invariance properties and by using molecular symmetry.
Semiempirical methods There are clearly computational limitations to treating molecular systems with large numbers of electrons accurately. Even with increases in computer speed and memory and the development of efficient algorithms, ab initio methods are not applied routinely to molecules with several dozen atoms. On the other hand, semiempirical methods are fast enough to be applied routinely to larger systems and, thus, make electronic structure calculations available for a wider range of molecules. Ab initio methods represent a more theoretically ‘pure’ approach, and one of the limitations to the accuracy of the semiempirical methods in addition to the approximations inherent in their formulation is the accuracy of experimental data used to obtain the parameters. However, in large part because adjustable parameters are optimized to reproduce a number of important chemical properties, semiempirical methods have become widely popular. The optimization of parameters is, in general, a difficult task for several reasons. First, accurate experimental data are often not available. Second, the simultaneous optimization of several parameters for a large number of molecules is very time-consuming. The parameters are interconnected in the sense that a significant variation in the value of one parameter in a nearly optimal parameter set must be accompanied by variations in several other parameters too. Successively optimizing each parameter is not feasible. Semiempirical methods were first developed for conjugated p-electron systems and we shall therefore begin our discussion with them and later describe more general methods.
326
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
9.17 Conjugated p-electron systems Consider the case of a conjugated p-system with a total of np p-electrons. The p-electrons are treated separately from the s-electrons partly because their energies are so different and partly on account of the different symmetries of their orbitals. The effective p-electron hamiltonian Hp is given by Hp ¼
np np np X X 2 X h 1 r2i þ Vip;eff þ 12j0 r 2me i¼1 i;j ij i¼1
ð9:53Þ
where the first term is the kinetic energy operator for the p-electrons, Vip;eff is the effective potential energy for p-electron i resulting from the potential field of the nuclei and all s-electrons, and the final term represents the repulsive potential energy due to interactions betwen p-electrons. The core hamiltonian hip for p-electron i is defined by hpi ¼
2 2 h r þ Vip;eff 2me i
ð9:54Þ
so we can write Hp ¼
np X
hpi þ 12j0
i¼1
np X 1 r i;j ij
ð9:55Þ
The analogy with eqn 7.46 should be apparent. The hamiltonian Hp is approximate because the p- and s-electrons have been treated separately and the effect of the latter in Hp appears only in the effective potential Vip;eff . The use of an approximate form for the hamiltonian of eqn 9.2 is characteristic of semiemipirical methods. The most famous semiempirical p-electron theory is the Hu¨ckel molecular orbital theory (HMO). As this method has already been described in some detail in Section 8.9, here we point out only some of the features of this method that characterize it as semiempirical. In the HMO method, Hp is approximated as a sum of one-electron terms: Hp ¼
np X
hp;eff i
ð9:56Þ
i¼1 p, eff is an effective hamiltonian for p-electron i. The form of hi is where hp;eff i left unspecified; only its matrix elements appear in HMO. Because Hp is a sum of one-electron terms, the wavefunction Cp can be written as a product of one-electron (molecular) orbitals ci, each of which is a solution of the eigenvalue equation
hip;eff ci ¼ Ei ci
ð9:57Þ
where Ei is the energy associated with the molecular orbital labelled i. Each molecular orbital is written as a linear combination of atomic orbitals (LCAO). For example, in an HMO treatment of a conjugated hydrocarbon (such as benzene), the molecular orbitals are linear combinations of C2pz
9.17 CONJUGATED p-ELECTRON SYSTEMS
j
327
atomic orbitals. The variation principle is then applied, and gives rise to a set of secular equations, which have non-trivial solutions only if detjhp;eff ESj ¼ 0
ð9:58Þ
is the matrix element of hp,eff between the atomic orbitals of the where hp;eff rs conjugated atoms r and s and Srs is their overlap integral. This expression is the analogue of the Roothaan equations, eqn 9.21, but differs in the restriction of the hamiltonian to a sum of one-electron terms and the orbitals to p-orbitals. Solution of the secular determinant yields the set of molecular orbital energies Ei as well as the expansion coefficients of the LCAO. As described in Section 8.9, HMO makes some severe assumptions about the and Srs: values of the matrix elements hp;eff rs 1. For all overlap integrals, Srs ¼ drs. p;eff ¼ a. 2. Diagonal elements hrr p;eff ¼ b if atoms r and s are neighbours and 0 3. Off-diagonal elements hrs otherwise. The setting of selected matrix elements to zero and the parametrizing of non-zero matrix elements are common features in semiempirical methods. Because Hp is written as a sum of one-electron terms with explicit forms left unspecified, the HMO method treats repulsions between the p-electrons very poorly (if at all!). As a result, it is useful only for qualitative discussions of p-conjugated systems. The Pariser–Parr–Pople method (PPP) is a much more substantial procedure than HMO, but nevertheless it is quite primitive when compared with current semiempirical procedures. It starts with the hamiltonian Hp of eqn 9.55 and writes the p-electron wavefunction Cp as a Slater determinant of p-electron spinorbitals fpi : Cp ¼ kfpa ð1Þfpb ð2Þ . . . fpz ðnp Þk
ð9:59Þ
The optimal spinorbitals are determined by using the variation principle, and satisfy f1p fpa ð1Þ ¼ epa fpa ð1Þ where epa is the orbital energy of spinorbital fpa and X f1p ¼ hp1 þ f Ju ð1Þ Ku ð1Þg
ð9:60Þ
ð9:61Þ
u
The Coulomb (Ju) and exchange (Ku) operators are defined as in eqn 9.10. At this stage, the calculation is following the ab initio route described in Sections 9.1 and 9.2. Indeed, proceeding as in Section 9.4 for the closed-shell case, we can write the p-electron spinorbital as a product of a spin function and a space function and expand the latter in a basis of known functions. The space functions are the p molecular orbitals and the basis functions are atomic orbitals yi centred on each p-conjugated atom i. For example, in a conjugated hydrocarbon, a basis set consisting of C2pz atomic orbitals is typically employed. The use of the basis set results in a set of equations
328
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
analogous to the Roothaan equations (eqn 9.19), with Fij in the latter replaced by the matrix elements Fijp : Z Fijp ¼ y i ð1Þf1p yj ð1Þ dr 1 ð9:62Þ and with ea replaced by epa . We simplify the notation by defining the one-electron integral Z ð9:63Þ hpij ¼ y i ð1Þhp1 yj ð1Þdr 1 and using the notation of eqn 9.24 for the two-electron integrals, we can write the matrix element Fijp as X Plm fðijjlmÞ 12ðimjljÞg ð9:64Þ Fijp ¼ hpij þ l, m where Plm is defined in eqn 9.27. At this point the PPP method makes some approximations beyond the separation of p and s orbitals. First, we set Sij ¼ dij, as in HMO theory. Then we set some of the two-electron integrals Z 1
y ð2Þyd ð2Þ dr 1 dr 2 ðab j cdÞ ¼ j0 y a ð1Þyb ð1Þ r12 c to zero, but in a more subtle way than in HMO. The product yi (1)yj(1) with i 6¼ j is called a differential overlap (it is the integrand of an overlap integral, so can formally be obtained from an overlap integral by differentiation; hence the name). In the zero differential overlap approximation (ZDO approximation), the two-electron integral is set to zero unless a ¼ b and c ¼ d. In other words, we set the product of atomic orbitals y a ð1Þyb ð1Þ ¼ 0 if a 6¼ b
ð9:65Þ
As a result, the two-electron integrals are given by ðab j cdÞ ¼ dab dcd ðaajccÞ
ð9:66Þ
and the integral (aajcc), which could be computed theoretically, is treated as an empirical parameter. In the ZDO approximation, all three-centre and four-centre two-electron integrals are neglected. In addition, in the PPP method the integrals hpij are usually not calculated theoretically but instead are set to zero or are treated as empirical parameters. In particular, for atomic orbitals yi and yj centred on atoms i and j that are not bonded together, hpij is set to zero; for atomic orbitals centred on atoms that are bonded together, the matrix element is taken to be an empirical parameter bij that varies with the nature of the atoms i and j. The diagonal elements hpii are usually set to an empirical parameter ai. (Note the resemblance to HMO theory at this point.) If all the two-electron integrals (abjcd) are set to zero and the matrix elements hpij replaced by the matrix elements hp;eff ij , then the PPP method (an SCF treatment) ‘reduces’ to the HMO method (a non-SCF treatment).
9.18 NEGLECT OF DIFFERENTIAL OVERLAP
j
329
9.18 Neglect of differential overlap The development of semiempirical methods to treat general molecular systems has made significant progress due in large part to the efforts of J.A. Pople and M.J.S. Dewar and their co-workers. These methods explicitly treat valence electrons and the names of the various methods are suggestive of which two-electron integrals are set to zero in the treatment. We shall set up the general equations for the treatment of the valence electrons and describe some of these semiempirical methods without going into detail.42 One point to keep in mind in the following discussion is that (except for hydrogen) there will be several basis functions (atomic orbitals) on each atom; this was not the case for conjugated systems and it makes the bookkeeping of neglected and non-neglected two-electron integrals more complicated. Consider a closed-shell molecule with nV valence electrons. The valenceelectron hamiltonian HV is given by HV ¼
nV X
1 hV i þ 2 j0
i¼1
nV X 1 r i;j ij
ð9:67Þ
where hVi is the core hamiltonian for valence electron i given by hV i ¼
2 2 h r þ ViV;eff 2me i
ð9:68Þ
and ViV,eff is the effective potential energy for valence electron i resulting from the potential field of the nuclei and all of the inner-shell electrons. We proceed exactly as we did for PPP in Section 9.17, and so obtain a set of equations identical to those of eqns 9.59–9.64 but with all the quantities previously labelled p now labelled V. This procedure results in a set of equations analogous to the Roothaan equations (eqn 9.19) to be solved in a self-consistent fashion with Fij in eqn 9.26 replaced by X FijV ¼ hV Plm fðijjlmÞ 12ðimjljÞg ð9:69Þ ij þ l;m
In the most primitive approach, known as the complete neglect of differential overlap (CNDO), we use the zero differential overlap approximation and write ðijjlmÞ ¼ dij dlm ðiijllÞ
ð9:70Þ
The two-electron integral is set to zero even when different atomic orbitals yi and yj belong to the same atom. The surviving integrals are often taken to be parameters with values that are adjusted until the results of the CNDO calculations resemble those of HF-SCF minimal basis set calculations. To discuss the next level of approximation, which is not as drastic as CNDO, we need to introduce some terminology. The ‘exchange integral’ was defined in eqn 7.37 in terms of the spatial parts of the spinorbitals; here we shall refer to .......................................................................................................
42. For details, see J.A. Pople and D.L. Beveridge, Approximate molecular orbital theory, McGraw-Hill, New York (1970).
330
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
the two-electron integrals over basis functions yi as exchange integrals if they are of the form (ijjij) (which is equal to (ijjji) if the basis functions are real). In the level of approximation known as intermediate neglect of differential overlap (INDO), we retain exchange integrals (ijjij) for which atomic orbitals yi and yj belong to the same atom. These one-centre exchange integrals are important for explaining the splitting between electronic states that come from the same electronic configuration; thus INDO will give vastly improved results over CNDO when spectroscopic terms are of interest. We now ask which two-electron integrals are retained in INDO. Consider a diagonal element FiiV in eqn 9.69. The first integral in the summation (ijjlm) becomes (iijlm) and the only contribution comes from the integral with l ¼ m. This integral (iijll) would also be retained in CNDO. The second integral in the summation (imjlj) becomes (imjli), and there are two contributions to it. One contribution comes from i ¼ l ¼ m giving the integral (iijii); the other contribution comes from m ¼ l with the stipulation that atomic orbital m belongs to the same atom as atomic orbital i. This one-centre exchange integral (imjmi) would not be retained in CNDO. Example 9.7 Identifying non-zero two-electron integrals in INDO
What two-electron integrals should be retained in the off-diagonal elements of FijV if atomic orbitals i and j are centred on the same atom? Method. We need to examine eqn 9.69 and consider each term in the
summation separately. We retain one-centre exchange integrals (ijjij) in addition to those two-electron integrals (iijll) retained in CNDO. Answer. The two-electron integral (ijjlm) contributes when i ¼ l and j ¼ m, or i ¼ m and j ¼ l, giving one-centre exchange integrals. The second term in the summation (imjlj) contributes (1) when i ¼ m and j ¼ l, giving an integral (iijll) that is also retained in CNDO, or (2) when i ¼ l and m ¼ j, giving the onecentre two-electron integral (ijjij). Self-test 9.7. What two-electron integrals contribute to the off-diagonal
elements FijV when atomic orbitals i and j belong to different atoms? [First term (ijjlm) never contributes; second term (imjlj) contributes only when i ¼ m and l ¼ j.]
As in CNDO, parameters are chosen in INDO to give as close agreement as possible to the results of minimal basis set HF-SCF calculations. Thus, although CNDO and INDO give reasonable equilibrium geometries when compared to experiment, they give poor results (as do HF-SCF methods) when compared with experimental quantities such as standard enthalpies of formation. Dewar and his co-workers developed a variety of semiempirical methods with the aim of reproducing not HF-SCF wavefunctions but rather four gas-phase molecular properties, namely molecular geometries, enthalpies of formation, dipole moments, and ionization energies. The hope was to achieve this goal by careful selection (that is, optimization) of the values of the parameters.
9.18 NEGLECT OF DIFFERENTIAL OVERLAP
j
331
Dewar first used the INDO approach and produced several versions of a semiempirical method he termed modified intermediate neglect of differential overlap (MINDO). The first two versions were called MINDO/1 and MINDO/2; a much improved version, which has proved to be useful for studies of hydrocarbons, is MINDO/3.43 Another semiempirical method based on INDO was developed by M.C. Zerner and denoted ZINDO/1. Its parametrization was optimized for calculating energies and geometries of molecules containing first or second series d-metals; it does not fare as well for organic molecules. ZINDO/S has been designed for computing electronic spectroscopic properties of d-metal complexes. A much less severe approximation than INDO is the neglect of diatomic differential overlap (NDDO) in which only diatomic differential overlap is not retained: that is, the differential overlap yi (1)yj(1) is neglected only when the basis functions belong to different atoms. Therefore, in the NDDO formalism, we retain all one-centre two-electron integrals and not just the one-centre exchange integrals; for example, (imjli) is retained where i, l, m are different basis functions on the same atom as well as retaining two-centre integrals of the form (abjcd) where a and b are different orbitals on one atom and c and d are different orbitals on a second atom. The NDDO method was proposed by Pople in the mid-1960s; however, it was not until Dewar developed the modified neglect of differential overlap method (MNDO) based on the NDDO formalism that the latter became more widely used as a predictive semiempirical method. In general, the MNDO method gives substantially better agreement with experiment than does MINDO/3. For example, in a study of 138 closed-shell molecules,44 the mean absolute error in enthalpies of formation decreased from about 50 kJ mol1 in MINDO/3 to 30 kJ mol1 in MNDO. Molecules were deliberately included in the study that had presented difficulties in earlier calculations and therefore the reported errors are larger than they would have been for a randomly selected set of molecules. Similarly, in a treatment of 80 different molecules and a total of 228 bonds,44 the mean absolute error in equilibrium bond length decreased from 2.2 pm in MINDO/3 to 1.4 pm in MNDO. Although MNDO was a significant improvement over MINDO/3, there remained deficiencies in the MNDO method, in particular an inadequate description of systems with hydrogen bonds. In 1985, Dewar developed an improved version of MNDO called Austin model 1 (AM1; the procedure is named after the University of Texas at Austin), which overcomes the major weaknesses of MNDO without any significant increase in computing time. AM1 also provides more accurate enthalpies of formation and (through the application of Koopmans’ theorem, Section 7.15) ionization energies than MNDO. However, there are instances in which AM1 can result in the prediction of some very peculiar geometries for hydrogen bonds.
.......................................................................................................
43. For a discussion of the parametrization of MINDO/3, see R.C. Bingham, M.J.S. Dewar, and D.H. Lo, J. Am. Chem. Soc., 1285, 97 (1975) and Section 2.4.1 of D.M. Hirst, A computational approach to chemistry, Blackwell Scientific Publications, Oxford (1990). 44. M.J.S. Dewar and W. Thiel, J. Am. Chem. Soc., 4907, 99 (1977).
332
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
Semiempirical methods continue to be developed because there is always room for improvement of the parametrization scheme and the use of experimental data. A third parametrization, designated PM3, of the MNDO method (MNDO and AM1 being versions 1 and 2, respectively) has been developed,45 and in general has been found to give better bond lengths, ionization energies, and enthalpies of formation than the two other MNDO schemes. However, despite this progress, there are cases where PM3 is much worse than MNDO. The three semiempirical methods MNDO, AM1, and PM3 do not use d-orbitals in their basis sets and therefore do not give very accurate results for most d-metal compounds nor for those main group elements where, from ab initio studies, d-orbitals are known to be of importance. The MNDO formalism has been extended to d-orbitals and the resulting MNDO/d parametrization scheme46 is much better at predicting properties of organometallic compounds. Furthermore, the AM1 scheme has been extended to include d-orbitals in a very similar manner to the creation of MNDO/d. This extension of AM1, which is parametrized for molybdenum and denoted AM1/d,47 gives identical results to AM1 for main-group atoms. For a series of Mo compounds, AM1/d resulted in mean absolute errors in bond distances of 4.4 pm and in enthalpies of formation of 30 kJ mol1. Another recent parametrization scheme is PM5, which yields results that are more accurate by up to a factor of four compared to PM3 and AM1; for example, the average error in absolute enthalpies of formation compared to experiment fell from 70 kJ mol1 using AM1 and 85 kJ mol1 using PM3 to 20 kJ mol1 using PM5. Other recently developed semiempirical methods include MSINDO, NDDO/MC, and PM3/tm. It is important to keep in mind that because the agreement between theoretical and experimental results does not necessarily improve for all chemical systems as a new parametrization scheme is introduced, it is essential to compare the outcome of several methods.
Molecular mechanics For very large systems, as in biochemical applications, it is not computationally practicable to use solely quantum mechanical approaches to compute potential energies. For these applications, often a mixture of quantum mechanics and molecular mechanics is employed, the latter using potential functions from classical mechanics to compute the potential energy for a specified arrangement of atoms.
.......................................................................................................
45. J.J.P. Stewart, J. Comput. Chem., 209, 10 (1989); idem., 221, 10 (1989). 46. W. Thiel and A.A. Voityuk, J. Phys. Chem., 616, 100 (1996). 47. A.A. Voityuk and N. Ro¨sch, J. Phys. Chem. A, 4089, 104 (2000).
9.19 FORCE FIELDS
j
333
9.19 Force fields In molecular mechanics (MM), the electrons in the system under study are not considered explicitly but rather each atom (the atomic nucleus and the associated electrons of the atom) is treated as a single particle. Therefore, MM is not very useful for chemical problems that involve bond-breaking or bondforming since electronic effects are critical in such cases. Rather, MM is commonly used in large systems for predicting the potential energy of a particular molecular conformation (that is, arrangement of atoms). The absolute values of the potential energies are not particularly meaningful in such calculations; instead it is energy differences between conformations that are significant. The atomic particles are treated as charged spheres with diameters determined from experiment or theory and charges (or partial charges) taken from theory. The interactions between the atoms are based on models of springs and on other classical potentials. The total potential energy is typically taken to be the sum of the bond stretching energy Estr, the bending energy Ebend, the twisting (or torsion) energy Etor, and the energy of interaction between nonbonded atoms Enb. The last contribution includes van der Waals, steric, and electrostatic interactions between atoms not chemically bound. (Other energy contributions such as stretch–bend coupling interactions and special treatment of hydrogen bonding might also appear in MM.) The equations for the potential energy terms contain parameters and the specified set of equations and parameters is called the force field. Numerous force fields have been developed for MM, the most common ones for studies of proteins and nucleic acids being AMBER and CHARMM. A popular force field for small organic molecules is N.L. Allinger’s MM2,48 part of the MMx set of force fields which also includes MM3 and MM4. Typical expressions for the potential energy terms are: P ð9:71aÞ Estr ¼ 12kb ðr r0 Þ2 Ebend ¼ Etor ¼ Enb
P1 2 2ky ðy y0 Þ
P
Af1 þ cosðnt fÞg
( ) X Cij Dij qi qj j0 ¼ 6 þ 12 þ er rij rij rij i>j
ð9:71bÞ ð9:71cÞ ð9:71dÞ
The sum in eqn 9.71a is over all bonds in the molecule and the harmonic oscillator parameters kb (an empirical force constant) and r0 (the equilibrium bond length) are assigned for each kind of bond (C–C, C–H, N–H, . . . ). The sum in eqn 9.71b extends over all angles and the values of ky (which controls the stiffness of the ‘angular’ spring) and y0 (the equilibrium angle) are assigned to each kind of angle (CCC, OCH, COH, . . . ). The sum in eqn 9.71c is over all torsional motions with A the amplitude, n the periodicity factor, .......................................................................................................
48. N.L. Allinger, J. Am. Chem. Soc., 8127, 99 (1977).
334
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
and f the displacement in the rotation angle t. The rotation angle is usually taken to be zero for the cis conformation of the quartet of atoms. The value of n reflects the symmetry of the torsional motion (for example, the HC–CH torsional motion in ethane has n ¼ 3 and periodicity 2p/3). Values of A, n, and f are assigned to each kind of torsional motion (CC–CC, CO–CC, CC–CN, . . . ). The sum in eqn 9.71d extends over all distinct pairs of interacting non-bonded atoms i and j. The first two terms in the equation represent van der Waals and repulsion interactions parametrized by empirical coefficients Cij and Dij which are assigned to each kind of non-bonded pair (C and H, C and C, C and O, . . . ). The last term in the sum is the Coulombic interaction of particles of charges qie and qje in an environment of local relative permittivity er. These charges (or partial charges) are often computed using ab initio or semiempirical methods. Because a given atom is surrounded by a small number of bonded atoms but can have non-bonded interactions with many atoms in the molecule, most of the computer time in MM is spent on computing Enb, and performing calculations on parallel and vector machines has been invaluable. The MM procedure is used to compute the potential energy for systems with large numbers, often thousands, of atoms, including biomolecules (such as proteins and nucleic acids), organic compounds, and polymers; both gasphase and solution-phase systems can be studied. The energy is computed for numerous molecular geometries so that the lowest energy conformation for the system can be located. Finding conformations corresponding to local minima in energy is a relatively easy task; determining the conformation associated with the global minimum is often a challenging undertaking due to the large number of degrees of freedom in macromolecules.
9.20 Quantum mechanics–molecular mechanics Calculations using MM take significantly less computer time than quantum mechanical methods; however, because the former do not provide descriptions in terms of electrons or orbitals, they are of limited use in systems where quantum mechanical effects are important. Therefore, an active area of current research efforts is the development of approaches that treat certain parts of the system accurately (that is, quantum mechanically) while treating other parts of the system with much faster methods of lower accuracy. One such approach is quantum mechanics–molecular mechanics (QM/MM). The QM/MM procedure is applicable when the system can be partitioned into two regions; one region (the ‘active site’) requires an accurate QM calculation of its potential and the second region (the rest of the system) acts as a perturbation on the active site and can be treated with an approximate and fast MM calculation of its potential. By using a quantum mechanical calculation, we can treat bond-breaking and bond-forming accurately at the active site yet still take into account the role of the surrounding atoms using MM. A QM/MM calculation is usually performed in an iterative manner. The conformation of the atoms in the active site (the QM subsystem) is fixed and the MM calculation is performed on the MM subsystem; in this latter calculation, the effects (for example, non-bonded interactions Enb) of the QM
9.20 QUANTUM MECHANICS–MOLECULAR MECHANICS
j
335
subsystem on the MM subsystem potential are taken into account by treating the QM atoms as fixed MM atoms. Then the QM method is applied to the active site utilizing, for example, a self-consistent field calculation that includes the potential energy of the MM subsystem. From the QM calculation, a preferred (lower energy) geometry of the QM subsystem is located. Then with this new active site conformation, the entire process is repeated until the geometry of all atoms in the system is converged. The critical challenge in a QM/MM calculation is to devise an accurate interface between QM and MM when the boundary between the two subsystems intersects covalent bonds. Various approaches have been developed in this active area of research;49 one common approach employs ‘capping’ atoms. As an example, suppose we are interested in an enzyme that contains a nitrogen atom in its active site which is bonded to a substituted benzene ring. If the QM/MM boundary is taken as intersecting the nitrogen–carbon covalent bond, the nitrogen atom is part of the QM subsystem and the substituted ring is part of the MM subsystem. In such an arrangement, the benzene ‘ligand’ is ‘removed’ and replaced with a capping atom, typically hydrogen, and the QM calculation is performed on the model subsystem with an N–H bond in the active site. The QM/MM procedure was first introduced in 1976 by Warshel and Levitt50 but a practical implementation was not given until a decade later by Singh and Kollman.51 One QM/MM scheme is integrated molecular orbital þ molecular mechanics (IMOMM), which was developed by Morokuma and co-workers and combines a high-level molecular orbital calculation with an MM calculation.52 The QM/MM procedure has been applied to a variety of large systems including nucleic acids, proteins, metalloenzymes and organometallic, catalysts. Three recent biomolecular applications are described briefly below. (1) Much of the ATP in the cells of living organisms is produced by the enzyme F1F0-ATP synthase. The F1 part of the enzyme, which is water soluble, can also catalyse the hydrolysis reaction converting ATP to ADP. A QM/MM investigation53 of the ATP hydrolysis occurring in the bTP site of F1 found the reaction to be endothermic, suggesting that the bTP site supports ATP synthesis rather than ATP hydrolysis in the specific enzyme conformation studied. (2) Cytochrome P-450 is an enzyme critically important in metabolism and modelling this (iron centre) enzyme is very useful to the pharmaceutical industry. The minimum-energy conformations of two cytochrome
.......................................................................................................
49. A discussion of some of these approaches can be found in Chapter 13 of C.J. Cramer, Essentials of computational chemistry: theories and models, Wiley & Sons, Chichester (2002). 50. A. Warshel and M. Levitt, J. Mol. Biol., 227, 103 (1976). 51. U.C. Singh and P.A. Kollman, J. Comput. Chem., 718, 7 (1986). 52. IMOMM is a subset of the ONIOM method (‘Our own N-layered Integrated molecular Orbital molecular Mechanics method’) described in R.D.J. Froese and K. Morokuma, Chem. Phys. Lett., 419, 305 (1999). 53. M. Dittrich, S. Hayashi, and K. Schulten, Biophys. J., 2253, 85 (2003).
336
j
9 THE CALCULATION OF ELECTRONIC STRUCTURE
P-450 containing species were located and the relative energies of doublet and quartet spin states determined in a QM/MM study employing DFT.54 (3) An important process in glucose fermentation is the enolase-catalysed proton transfer of 2-phospho-D-glycerate (2-PGA) to phosphoenolpyruvate. The potential energy surface for the reaction was computed using QM/MM.55 The QM subsystem consisted of 2-PGA and a ten-atom portion of the Lys345 side chain of the enolase enzyme (the Lys345 enzyme residue having been identified in previous activity assays as the general catalyst for proton abstraction). The 25-atom QM subsystem was characterized using AM1; the MM calculation included a CHARMM force field.
Software packages for electronic structure calculations Sophisticated software packages have been developed over the past four decades to perform electronic structure calculations using the methods we have described. These packages are widely available and are becoming increasingly easy to use. They are often accompanied by sophisticated graphical user interfaces (GUIs) that allow visualization of results of calculations. Many packages can be used to compute both potential energies and analytic derivatives, and so are useful for determining equilibrium geometries, transition state geometries, and vibrational frequencies. We mention here some of the available packages and emphasize methods we have discussed in this chapter. See the Further reading section for a more complete discussion. Several widely used software packages capable of electronic structure calculations include Gaussian, GAMESS (General Atomic and Molecular Electronic Structure System), and CADPAC (Cambridge Analytical Derivatives Package). Gaussian utilizes a wide variety of ab initio, DFT, semiempirical, and MM methods (including solvation models). GAMESS (and its United Kingdom counterpart GAMESS-UK) uses CI, MP2, CC, and DFT for electron correlation corrections; it also has semiempirical capabilities. CADPAC (part of the UniChem software package) is capable of HF, MPPT, CC, and CI calculations. Other quantum mechanical software packages are ACES-II (with strengths in MBPT, CC, and DFT calculations) and Dalton QCP (Quantum Chemistry Program), which can compute electric and magnetic properties using SCF, MP2, CC, and MCSCF. Several packages designed for highly accurate computations on relatively small molecules by focusing on accurate multiconfiguration and multireference treatments of .......................................................................................................
54. R.B. Murphy, D.M. Philipp, and R.A. Friesner, J. Comput. Chem., 1442, 21 (2000). 55. C. Alhambra, J. Gao, J.C. Corchado, J. Villa`, and D.G. Truhlar, J. Am. Chem. Soc., 2253, 121 (1999).
SOFTWARE PACKAGES FOR ELECTRONIC STRUCTURE CALCULATIONS
j
337
electron correlation are MOLCAS, MOLPRO, and COLUMBUS. MELDF uses MRCI and multireference MP2 methods. NWChem and MPQC (Massively Parallel Quantum Chemistry Code) are designed for peak efficiency on machines capable of parallel computing. Other ab initio software packages are Jaguar (which can treat large systems such as molecular clusters), Q-Chem (which treats solvation using the CHEMSOL package), Turbomole (which uses Gaussian basis sets), and HONDO (a semiempirical molecular orbital program with SCF, MCSCF, and CI wavefunctions). Software packages designed for density functional theory calculations include deMon, DMol, DGauss, and ADF (Amsterdam Density Functional). The last is popular for work in catalysis, spectroscopy, inorganic chemistry, biochemistry (using QM/MM), and heavy elements (using relativistic methods). DGauss, part of UniChem, uses Gaussian-type orbitals for computational efficiency. The DMol program can handle molecular clusters and periodic systems. Two widely used semiempirical packages are MOPAC and AMPAC. MOPAC, which is the most popular semiempirical software package, includes MNDO, MINDO/3, AM1, PM3, MNDO/d, and PM5 parametrization schemes. It can characterize properties and reactivity of large systems (hundreds of atoms) in the gas phase and in solution as well as in the solid state. AMPAC and the UniChem software program called MNDO also include a variety of semiempirical methodologies. Other packages include VAMP, which is designed for organic and inorganic systems, and ZINDO, which uses a parametrization for molecular spectroscopic properties. The semiempirical package AMSOL includes a variety of solvation models (primarily based on the AM1 and PM3 methods) for computing Gibbs energies of solvation in water and numerous organic solvents. Be aware, however, that because the semiempirical methodologies are parametrization-dependent, AM1 in one software package, for example, need not be the same as AM1 in another software package. Software packages for molecular mechanics include CHARMM (Chemistry at Harvard Macromolecular Mechanics), AMBER (Assisted Model Building with Energy Refinement), SIFBA (Sum of Interactions Between Fragments Ab initio computed), QuanteMM, and TINKER. CHARMM uses potential energy functions parametrized for proteins, nucleic acids (DNA and RNA), and lipids. Energies can be evaluated once parameters (such as force constants, equilibrium geometries, and van der Waals radii) are specified; the values of parameters are obtained from a combination of experimental and quantum mechanical studies. The commercial version of CHARMM, called CHARMm, is capable of QM/MM computations by interfacing to several different QM programs. SIFBA focuses on inter- and intramolecular interactions of organic and biological molecules. QuanteMM is capable of studying a diverse range of systems including biomolecules and inorganic catalysts. It undertakes QM/MM studies by using DMol, Turbomole, or MOPAC on one part of the system and molecular mechanics on the ‘surrounding’ atoms. AMBER provides efficient and accurate biomolecular force fields for MM. TINKER uses numerous force fields including AMBER and CHARMM parameter sets; it has special features for biopolymers.
